Mastering the Poisson Distribution: Intuition and Foundations

Stay Ahead, Stay ONMINE

Mastering the Poisson Distribution: Intuition and Foundations

You’ve probably used the normal distribution one or two times too many. We all have — It’s a true workhorse. But sometimes, we run into problems. For instance, when predicting or forecasting values, simulating data given a particular data-generating process, or when we try to visualise model output and explain them intuitively to non-technical stakeholders. Suddenly, things don’t make much sense: can a user really have made -8 clicks on the banner? Or even 4.3 clicks? Both are examples of how count data doesn’t behave. I’ve found that better encapsulating the data generating process into my modelling has been key to having sensible model output. Using the Poisson distribution when it was appropriate has not only helped me convey more meaningful insights to stakeholders, but it has also enabled me to produce more accurate error estimates, better Inference, and sound decision-making. In this post, my aim is to help you get a deep intuitive feel for the Poisson distribution by walking through example applications, and taking a dive into the foundations — the maths. I hope you learn not just how it works, but also why it works, and when to apply the distribution. If you know of a resource that has helped you grasp the concepts in this blog particularly well, you’re invited to share it in the comments! Outline Examples and use cases: Let’s walk through some use cases and sharpen the intuition I just mentioned. Along the way, the relevance of the Poisson Distribution will become clear. The foundations: Next, let’s break down the equation into its individual components. By studying each part, we’ll uncover why the distribution works the way it does. The assumptions: Equipped with some formality, it will be easier to understand the assumptions that power the distribution, and at the same time set the boundaries for when it works, and when not. When real life deviates from the model: Finally, let’s explore the special links that the Poisson distribution has with the Negative Binomial distribution. Understanding these relationships can deepen our understanding, and provide alternatives when the Poisson distribution is not suited for the job. Example in an online marketplace I chose to deep dive into the Poisson distribution because it frequently appears in my day-to-day work. Online marketplaces rely on binary user choices from two sides: a seller deciding to list an item and a buyer deciding to make a purchase. These micro-behaviours drive supply and demand, both in the short and long term. A marketplace is born. Binary choices aggregate into counts — the sum of many such decisions as they occur. Attach a timeframe to this counting process, and you’ll start seeing Poisson distributions everywhere. Let’s explore a concrete example next. Consider a seller on a platform. In a given month, the seller may or may not list an item for sale (a binary choice). We would only know if she did because then we’d have a measurable count of the event. Nothing stops her from listing another item in the same month. If she does, we count those events. The total could be zero for an inactive seller or, say, 120 for a highly engaged seller. Over several months, we would observe a varying number of listed items by this seller — sometimes fewer, sometimes more — hovering around an average monthly listing rate. That is essentially a Poisson process. When we get to the assumptions section, you’ll see what we had to assume away to make this example work. Other examples Other phenomena that can be modelled with a Poisson distribution include: Sports analytics: The number of goals scored in a match between two teams. Queuing: Customers arriving at a help desk or customer support calls. Insurance: The number of claims made within a given period. Each of these examples warrants further inspection, but for the remainder of this post, we’ll use the marketplace example to illustrate the inner workings of the distribution. The mathy bit … or foundations. I find opening up the probability mass function (PMF) of distributions helpful to understanding why things work as they do. The PMF of the Poisson distribution goes like: Where λ is the rate parameter, and 𝑘 is the manifested count of the random variable (𝑘 = 0, 1, 2, 3, … events). Very neat and compact. The probability mass function of the Poisson distribution, for a few different lambdas. Contextualising λ and k: the marketplace example In the context of our earlier example — a seller listing items on our platform — λ represents the seller’s average monthly listings. As the expected monthly value for this seller, λ orchestrates the number of items she would list in a month. Note that λ is a Greek letter, so read: λ is a parameter that we can estimate from data. On the other hand, 𝑘 does not hold any information about the seller’s idiosyncratic behaviour. It’s the target value we set for the number of events that may happen to learn about its probability. The dual role of λ as the mean and variance When I said that λ orchestrates the number of monthly listings for the seller, I meant it quite literally. Namely, λ is both the expected value and variance of the distribution, indifferently, for all values of λ. This means that the mean-to-variance ratio (index of dispersion) is always 1. To put this into perspective, the normal distribution requires two parameters — 𝜇 and 𝜎², the average and variance respectively — to fully describe it. The Poisson distribution achieves the same with just one. Having to estimate only one parameter can be beneficial for parametric inference. Specifically, by reducing the variance of the model and increasing the statistical power. On the other hand, it can be too limiting of an assumption. Alternatives like the Negative Binomial distribution can alleviate this limitation. We’ll explore that later. Breaking down the probability mass function Now that we know the smallest building blocks, let’s zoom out one step: what is λᵏ, 𝑒^⁻λ, and 𝑘!, and more importantly, what is each of these components’ function in the whole? λᵏ is a weight that expresses how likely it is for 𝑘 events to happen, given that the expectation is λ. Note that “likely” here does not mean a probability, yet. It’s merely a signal strength. 𝑘! is a combinatorial correction so that we can say that the order of the events is irrelevant. The events are interchangeable. 𝑒^⁻λ normalises the integral of the PMF function to sum up to 1. It’s called the partition function of exponential-family distributions. In more detail, λᵏ relates the observed value 𝑘 to the expected value of the random variable, λ. Intuitively, more probability mass lies around the expected value. Hence, if the observed value lies close to the expectation, the probability of occurring is larger than the probability of an observation far removed from the expectation. Before we can cross-check our intuition with the numerical behaviour of λᵏ, we need to consider what 𝑘! does. Interchangeable events Had we cared about the order of events, then each unique event could be ordered in 𝑘! ways. But because we don’t, and we deem each event interchangeable, we “divide out” 𝑘! from λᵏ to correct for the overcounting. Since λᵏ is an exponential term, the output will always be larger as 𝑘 grows, holding λ constant. That is the opposite of our intuition that there is maximum probability when λ = 𝑘, as the output is larger when 𝑘 = λ + 1. But now that we know about the interchangeable events assumption — and the overcounting issue — we know that we have to factor in 𝑘! like so: λᵏ 𝑒^⁻λ / 𝑘!, to see the behaviour we expect. Now let’s check the intuition of the relationship between λ and 𝑘 through λᵏ, corrected for 𝑘!. For the same λ, say λ = 4, we should see λᵏ 𝑒^⁻λ / 𝑘! to be smaller for values of 𝑘 that are far removed from 4, compared to values of 𝑘 that lie close to 4. Like so: inline code: 4²/2 = 8 is smaller than 4⁴/24 = 10.7. This is consistent with the intuition of a higher likelihood of 𝑘 when it’s near the expectation. The image below shows this relationship more generally, where you see that the output is larger as 𝑘 approaches λ. The probability mass function without the normalising component e^-lambda. The assumptions First, let’s get one thing off the table: the difference between a Poisson process, and the Poisson distribution. The process is a stochastic continuous-time model of points happening in given interval: 1D, a line; 2D, an area, or higher dimensions. We, data scientists, most often deal with the one-dimensional case, where the “line” is time, and the points are the events of interest — I dare to say. These are the assumptions of the Poisson process: The occurrence of one event does not affect the probability of a second event. Think of our seller going on to list another item tomorrow indifferently of having done so already today, or the one from five days ago for that matter. The point here is that there is no memory between events. The average rate at which events occur, is independent of any occurrence. In other words, no event that happened (or will happen) alters λ, which remains constant throughout the observed timeframe. In our seller example, this means that listing an item today does not increase or decrease the seller’s motivation or likelihood of listing another item tomorrow. Two events cannot occur at exactly the same instant. If we were to zoom at an infinite granular level on the timescale, no two listings could have been placed simultaneously; always sequentially. From these assumptions — no memory, constant rate, events happening alone — it follows that 1) any interval’s number of events is Poisson-distributed with parameter λₜ and 2) that disjoint intervals are independent — two key properties of a Poisson process. A Note on the distribution:The distribution simply describes probabilities for various numbers of counts in an interval. Strictly speaking, one can use the distribution pragmatically whenever the data is nonnegative, can be unbounded on the right, has mean λ, and reasonably models the data. It would be just convenient if the underlying process is a Poisson one, and actually justifies using the distribution. The marketplace example: Implications So, can we justify using the Poisson distribution for our marketplace example? Let’s open up the assumptions of a Poisson process and take the test. Constant λ Why it may fail: The seller has patterned online activity; holidays; promotions; listings are seasonal goods. Consequence: λ is not constant, leading to overdispersion (mean-to-variance ratio is larger than 1, or to temporal patterns. Independence and memorylessness Why it may fail: The propensity to list again is higher after a successful listing, or conversely, listing once depletes the stock and intervenes with the propensity of listing again. Consequence: Two events are no longer independent, as the occurrence of one informs the occurrence of the other. Simultaneous events Why it may fail: Batch-listing, a new feature, was introduced to help the sellers. Consequence: Multiple listings would come online at the same time, clumped together, and they would be counted simultaneously. Balancing rigour and pragmatism As Data Scientists on the job, we may feel trapped between rigour and pragmatism. The three steps below should give you a sound foundation to decide on which side to err, when the Poisson distribution falls short: Pinpoint your goal: is it inference, simulation or prediction, and is it about high-stakes output? List the worst thing that can happen, and the cost of it for the business. Identify the problem and solution: why does the Poisson distribution not fit, and what can you do about it? list 2-3 solutions, including changing nothing. Balance gains and costs: Will your workaround improve things, or make it worse? and at what cost: interpretability, new assumptions introduced and resources used. Does it help you in achieving your goal? That said, here are some counters I use when needed. When real life deviates from your model Everything described so far pertains to the standard, or homogenous, Poisson process. But what if reality begs for something different? In the next section, we’ll cover two extensions of the Poisson distribution when the constant λ assumption does not hold. These are not mutually exclusive, but neither they are the same: Time-varying λ: a single seller whose listing rate ramps up before holidays and slows down afterward Mixed Poisson distribution: multiple sellers listing items, each with their own λ can be seen as a mixture of various Poisson processes Time-varying λ The first extension allows λ to have its own value for each time t. The PMF then becomes Where the number of events 𝐾(𝑇) in an interval 𝑇 follows the Poisson distribution with a rate no longer equal to a fixed λ, but one equal to: More intuitively, integrating over the interval 𝑡 to 𝑡 + 𝑖 gives us a single number: the expected value of events over that interval. The integral will vary by each arbitrary interval, and that’s what makes λ change over time. To understand how that integration works, it was helpful for me to think of it like this: if the interval 𝑡 to 𝑡₁ integrates to 3, and 𝑡₁ to 𝑡₂ integrates to 5, then the interval 𝑡 to 𝑡₂ integrates to 8 = 3 + 5. That’s the two expectations summed up, and now the expectation of the entire interval. Practical implication One may want to modeling the expected value of the Poisson distribution as a function of time. For instance, to model an overall change in trend, or seasonality. In generative model notation: Time may be a continuous variable, or an arbitrary function of it. Process-varying λ: Mixed Poisson distribution But then there’s a gotcha. Remember when I said that λ has a dual role as the mean and variance? That still applies here. Looking at the “relaxed” PMF*, the only thing that changes is that λ can vary freely with time. But it’s still the one and only λ that orchestrates both the expected value and the dispersion of the PMF*. More precisely, 𝔼[𝑋] = Var(𝑋) still holds. There are various reasons for this constraint not to hold in reality. Model misspecification, event interdependence and unaccounted for heterogeneity could be the issues at hand. I’d like to focus on the latter case, as it justifies the Negative Binomial distribution — one of the topics I promised to open up. Heterogeneity and overdispersionImagine we are not dealing with one seller, but with 10 of them listing at different intensity levels, λᵢ, where 𝑖 = 1, 2, 3, …, 10 sellers. Then, essentially, we have 10 Poisson processes going on. If we unify the processes and estimate the grand λ, we simplify the mixture away. Meaning, we get a correct estimate of all sellers on average, but the resulting grand λ is naive and does not know about the original spread of λᵢ. It still assumes that the variance and mean are equal, as per the axioms of the distribution. This will lead to overdispersion and, in turn, to underestimated errors. Ultimately, it inflates the false positive rate and drives poor decision-making. We need a way to embrace the heterogeneity amongst sellers’ λᵢ. Negative binomial: Extending the Poisson distributionAmong the few ways one can look at the Negative Binomial distribution, one way is to see it as a compound Poisson process — 10 sellers, sounds familiar yet? That means multiple independent Poisson processes are summed up to a single one. Mathematically, first we draw λ from a Gamma distribution: λ ~ Γ(r, θ), then we draw the count 𝑋 | λ ~ Poisson(λ). In one image, it is as if we would sample from plenty Poisson distributions, corresponding to each seller. A negative Binomial distribution arises from many Poisson distributions. The more exposing alias of the Negative binomial distribution is Gamma-Poisson mixture distribution, and now we know why: the dictating λ comes from a continuous mixture. That’s what we needed to explain the heterogeneity amongst sellers. Let’s simulate this scenario to gain more intuition. Gamma mixture of lambda. First, we draw λᵢ from a Gamma distribution: λᵢ ~ Γ(r, θ). Intuitively, the Gamma distribution tells us about the variety in the intensity — listing rate — amongst the sellers. On a practical note, one can instill their assumptions about the degree of heterogeneity in this step of the model: how different are sellers? By varying the levels of heterogeneity, one can observe the impact on the final Poisson-like distribution. Doing this type of checks (i.e., posterior predictive check), is common in Bayesian modeling, where the assumptions are set explicitly. Gamma-Poisson mixture distribution versus homogenous Poisson distribution. Τhe dashed line reflects λ, which is 4 for both distributions. In the second step, we plug the obtained λ into the Poisson distribution: 𝑋 | λ ~ Poisson(λ), and obtain a Poisson-like distribution that represents the summed subprocesses. Notably, this unified process has a larger dispersion than expected from a homogeneous Poisson distribution, but it is in line with the Gamma mixture of λ. Heterogeneous λ and inference A practical consequence of introducing flexibility into your assumed distribution is that inference becomes more challenging. More parameters (i.e., the Gamma parameters) need to be estimated. Parameters act as flexible explainers of the data, tending to overfit and explain away variance in your variable. The more parameters you have, the better the explanation may seem, but the model also becomes more susceptible to noise in the data. Higher variance reduces the power to identify a difference in means, if one exists, because — well — it gets lost in the variance. Countering the loss of power Confirm whether you indeed need to extend the standard Poisson distribution. If not, simplify to the simplest, most fit model. A quick check on overdispersion may suffice for this. Pin down the estimates of the Gamma mixture distribution parameters using regularising, informative priors (think: Bayes). During my research process for writing this blog, I learned a great deal about the connective tissue underlying all of this: how the binomial distribution plays a fundamental role in the processes we’ve discussed. And while I’d love to ramble on about this, I’ll save it for another post, perhaps. In the meantime, feel free to share your understanding in the comments section below 👍. Conclusion The Poisson distribution is a simple distribution that can be highly suitable for modelling count data. However, when the assumptions do not hold, one can extend the distribution by allowing the rate parameter to vary as a function of time or other factors, or by assuming subprocesses that collectively make up the count data. This added flexibility can address the limitations, but it comes at a cost: increased flexibility in your modelling raises the variance and, consequently, undermines the statistical power of your model. If your end goal is inference, you may want to think twice and consider exploring simpler models for the data. Alternatively, switch to the Bayesian paradigm and leverage its built-in solution to regularise estimates: informative priors. I hope this has given you what you came for — a better intuition about the Poisson distribution. I’d love to hear your thoughts about this in the comments! Unless otherwise noted, all images are by the author.Originally published at https://aalvarezperez.github.io on January 5, 2025.

I’ve found that better encapsulating the data generating process into my modelling has been key to having sensible model output. Using the Poisson distribution when it was appropriate has not only helped me convey more meaningful insights to stakeholders, but it has also enabled me to produce more accurate error estimates, better Inference, and sound decision-making.

In this post, my aim is to help you get a deep intuitive feel for the Poisson distribution by walking through example applications, and taking a dive into the foundations — the maths. I hope you learn not just how it works, but also why it works, and when to apply the distribution.

If you know of a resource that has helped you grasp the concepts in this blog particularly well, you’re invited to share it in the comments!

Outline

Examples and use cases: Let’s walk through some use cases and sharpen the intuition I just mentioned. Along the way, the relevance of the Poisson Distribution will become clear.
The foundations: Next, let’s break down the equation into its individual components. By studying each part, we’ll uncover why the distribution works the way it does.
The assumptions: Equipped with some formality, it will be easier to understand the assumptions that power the distribution, and at the same time set the boundaries for when it works, and when not.
When real life deviates from the model: Finally, let’s explore the special links that the Poisson distribution has with the Negative Binomial distribution. Understanding these relationships can deepen our understanding, and provide alternatives when the Poisson distribution is not suited for the job.

Example in an online marketplace

I chose to deep dive into the Poisson distribution because it frequently appears in my day-to-day work. Online marketplaces rely on binary user choices from two sides: a seller deciding to list an item and a buyer deciding to make a purchase. These micro-behaviours drive supply and demand, both in the short and long term. A marketplace is born.

Binary choices aggregate into counts — the sum of many such decisions as they occur. Attach a timeframe to this counting process, and you’ll start seeing Poisson distributions everywhere. Let’s explore a concrete example next.

Consider a seller on a platform. In a given month, the seller may or may not list an item for sale (a binary choice). We would only know if she did because then we’d have a measurable count of the event. Nothing stops her from listing another item in the same month. If she does, we count those events. The total could be zero for an inactive seller or, say, 120 for a highly engaged seller.

Over several months, we would observe a varying number of listed items by this seller — sometimes fewer, sometimes more — hovering around an average monthly listing rate. That is essentially a Poisson process. When we get to the assumptions section, you’ll see what we had to assume away to make this example work.

Other examples

Other phenomena that can be modelled with a Poisson distribution include:

Sports analytics: The number of goals scored in a match between two teams.
Queuing: Customers arriving at a help desk or customer support calls.
Insurance: The number of claims made within a given period.

Each of these examples warrants further inspection, but for the remainder of this post, we’ll use the marketplace example to illustrate the inner workings of the distribution.

The mathy bit

… or foundations.

I find opening up the probability mass function (PMF) of distributions helpful to understanding why things work as they do. The PMF of the Poisson distribution goes like:

Where λ is the rate parameter, and 𝑘 is the manifested count of the random variable (𝑘 = 0, 1, 2, 3, … events). Very neat and compact.

Graph: The probability mass function of the Poisson distribution, for a few different lambdas. — The probability mass function of the Poisson distribution, for a few different lambdas.

Contextualising λ and k: the marketplace example

In the context of our earlier example — a seller listing items on our platform — λ represents the seller’s average monthly listings. As the expected monthly value for this seller, λ orchestrates the number of items she would list in a month. Note that λ is a Greek letter, so read: λ is a parameter that we can estimate from data. On the other hand, 𝑘 does not hold any information about the seller’s idiosyncratic behaviour. It’s the target value we set for the number of events that may happen to learn about its probability.

The dual role of λ as the mean and variance

When I said that λ orchestrates the number of monthly listings for the seller, I meant it quite literally. Namely, λ is both the expected value and variance of the distribution, indifferently, for all values of λ. This means that the mean-to-variance ratio (index of dispersion) is always 1.

To put this into perspective, the normal distribution requires two parameters — 𝜇 and 𝜎², the average and variance respectively — to fully describe it. The Poisson distribution achieves the same with just one.

Having to estimate only one parameter can be beneficial for parametric inference. Specifically, by reducing the variance of the model and increasing the statistical power. On the other hand, it can be too limiting of an assumption. Alternatives like the Negative Binomial distribution can alleviate this limitation. We’ll explore that later.

Breaking down the probability mass function

Now that we know the smallest building blocks, let’s zoom out one step: what is λᵏ, 𝑒^⁻λ, and 𝑘!, and more importantly, what is each of these components’ function in the whole?

λᵏ is a weight that expresses how likely it is for 𝑘 events to happen, given that the expectation is λ. Note that “likely” here does not mean a probability, yet. It’s merely a signal strength.
𝑘! is a combinatorial correction so that we can say that the order of the events is irrelevant. The events are interchangeable.
𝑒^⁻λ normalises the integral of the PMF function to sum up to 1. It’s called the partition function of exponential-family distributions.

In more detail, λᵏ relates the observed value 𝑘 to the expected value of the random variable, λ. Intuitively, more probability mass lies around the expected value. Hence, if the observed value lies close to the expectation, the probability of occurring is larger than the probability of an observation far removed from the expectation. Before we can cross-check our intuition with the numerical behaviour of λᵏ, we need to consider what 𝑘! does.

Interchangeable events

Had we cared about the order of events, then each unique event could be ordered in 𝑘! ways. But because we don’t, and we deem each event interchangeable, we “divide out” 𝑘! from λᵏ to correct for the overcounting.

Since λᵏ is an exponential term, the output will always be larger as 𝑘 grows, holding λ constant. That is the opposite of our intuition that there is maximum probability when λ = 𝑘, as the output is larger when 𝑘 = λ + 1. But now that we know about the interchangeable events assumption — and the overcounting issue — we know that we have to factor in 𝑘! like so: λᵏ 𝑒^⁻λ / 𝑘!, to see the behaviour we expect.

Now let’s check the intuition of the relationship between λ and 𝑘 through λᵏ, corrected for 𝑘!. For the same λ, say λ = 4, we should see λᵏ 𝑒^⁻λ / 𝑘! to be smaller for values of 𝑘 that are far removed from 4, compared to values of 𝑘 that lie close to 4. Like so: inline code: 4²/2 = 8 is smaller than 4⁴/24 = 10.7. This is consistent with the intuition of a higher likelihood of 𝑘 when it’s near the expectation. The image below shows this relationship more generally, where you see that the output is larger as 𝑘 approaches λ.

Graph: The probability mass function without the normalising component e^-lambda. — The probability mass function without the normalising component e^-lambda.

The assumptions

First, let’s get one thing off the table: the difference between a Poisson process, and the Poisson distribution. The process is a stochastic continuous-time model of points happening in given interval: 1D, a line; 2D, an area, or higher dimensions. We, data scientists, most often deal with the one-dimensional case, where the “line” is time, and the points are the events of interest — I dare to say.

These are the assumptions of the Poisson process:

The occurrence of one event does not affect the probability of a second event. Think of our seller going on to list another item tomorrow indifferently of having done so already today, or the one from five days ago for that matter. The point here is that there is no memory between events.
The average rate at which events occur, is independent of any occurrence. In other words, no event that happened (or will happen) alters λ, which remains constant throughout the observed timeframe. In our seller example, this means that listing an item today does not increase or decrease the seller’s motivation or likelihood of listing another item tomorrow.
Two events cannot occur at exactly the same instant. If we were to zoom at an infinite granular level on the timescale, no two listings could have been placed simultaneously; always sequentially.

From these assumptions — no memory, constant rate, events happening alone — it follows that 1) any interval’s number of events is Poisson-distributed with parameter λₜ and 2) that disjoint intervals are independent — two key properties of a Poisson process.

A Note on the distribution:
The distribution simply describes probabilities for various numbers of counts in an interval. Strictly speaking, one can use the distribution pragmatically whenever the data is nonnegative, can be unbounded on the right, has mean λ, and reasonably models the data. It would be just convenient if the underlying process is a Poisson one, and actually justifies using the distribution.

The marketplace example: Implications

So, can we justify using the Poisson distribution for our marketplace example? Let’s open up the assumptions of a Poisson process and take the test.

Constant λ

Why it may fail: The seller has patterned online activity; holidays; promotions; listings are seasonal goods.
Consequence: λ is not constant, leading to overdispersion (mean-to-variance ratio is larger than 1, or to temporal patterns.

Independence and memorylessness

Why it may fail: The propensity to list again is higher after a successful listing, or conversely, listing once depletes the stock and intervenes with the propensity of listing again.
Consequence: Two events are no longer independent, as the occurrence of one informs the occurrence of the other.

Simultaneous events

Why it may fail: Batch-listing, a new feature, was introduced to help the sellers.
Consequence: Multiple listings would come online at the same time, clumped together, and they would be counted simultaneously.

Balancing rigour and pragmatism

As Data Scientists on the job, we may feel trapped between rigour and pragmatism. The three steps below should give you a sound foundation to decide on which side to err, when the Poisson distribution falls short:

Pinpoint your goal: is it inference, simulation or prediction, and is it about high-stakes output? List the worst thing that can happen, and the cost of it for the business.
Identify the problem and solution: why does the Poisson distribution not fit, and what can you do about it? list 2-3 solutions, including changing nothing.
Balance gains and costs: Will your workaround improve things, or make it worse? and at what cost: interpretability, new assumptions introduced and resources used. Does it help you in achieving your goal?

That said, here are some counters I use when needed.

When real life deviates from your model

Everything described so far pertains to the standard, or homogenous, Poisson process. But what if reality begs for something different?

In the next section, we’ll cover two extensions of the Poisson distribution when the constant λ assumption does not hold. These are not mutually exclusive, but neither they are the same:

Time-varying λ: a single seller whose listing rate ramps up before holidays and slows down afterward
Mixed Poisson distribution: multiple sellers listing items, each with their own λ can be seen as a mixture of various Poisson processes

Time-varying λ

The first extension allows λ to have its own value for each time t. The PMF then becomes

Where the number of events 𝐾(𝑇) in an interval 𝑇 follows the Poisson distribution with a rate no longer equal to a fixed λ, but one equal to:

More intuitively, integrating over the interval 𝑡 to 𝑡 + 𝑖 gives us a single number: the expected value of events over that interval. The integral will vary by each arbitrary interval, and that’s what makes λ change over time. To understand how that integration works, it was helpful for me to think of it like this: if the interval 𝑡 to 𝑡₁ integrates to 3, and 𝑡₁ to 𝑡₂ integrates to 5, then the interval 𝑡 to 𝑡₂ integrates to 8 = 3 + 5. That’s the two expectations summed up, and now the expectation of the entire interval.

Practical implication
One may want to modeling the expected value of the Poisson distribution as a function of time. For instance, to model an overall change in trend, or seasonality. In generative model notation:

Time may be a continuous variable, or an arbitrary function of it.

Process-varying λ: Mixed Poisson distribution

But then there’s a gotcha. Remember when I said that λ has a dual role as the mean and variance? That still applies here. Looking at the “relaxed” PMF*, the only thing that changes is that λ can vary freely with time. But it’s still the one and only λ that orchestrates both the expected value and the dispersion of the PMF*. More precisely, 𝔼[𝑋] = Var(𝑋) still holds.

There are various reasons for this constraint not to hold in reality. Model misspecification, event interdependence and unaccounted for heterogeneity could be the issues at hand. I’d like to focus on the latter case, as it justifies the Negative Binomial distribution — one of the topics I promised to open up.

Heterogeneity and overdispersion
Imagine we are not dealing with one seller, but with 10 of them listing at different intensity levels, λᵢ, where 𝑖 = 1, 2, 3, …, 10 sellers. Then, essentially, we have 10 Poisson processes going on. If we unify the processes and estimate the grand λ, we simplify the mixture away. Meaning, we get a correct estimate of all sellers on average, but the resulting grand λ is naive and does not know about the original spread of λᵢ. It still assumes that the variance and mean are equal, as per the axioms of the distribution. This will lead to overdispersion and, in turn, to underestimated errors. Ultimately, it inflates the false positive rate and drives poor decision-making. We need a way to embrace the heterogeneity amongst sellers’ λᵢ.

Negative binomial: Extending the Poisson distribution
Among the few ways one can look at the Negative Binomial distribution, one way is to see it as a compound Poisson process — 10 sellers, sounds familiar yet? That means multiple independent Poisson processes are summed up to a single one. Mathematically, first we draw λ from a Gamma distribution: λ ~ Γ(r, θ), then we draw the count 𝑋 | λ ~ Poisson(λ).

In one image, it is as if we would sample from plenty Poisson distributions, corresponding to each seller.

A negative Binomial distribution arises from many Poisson distributions.

The more exposing alias of the Negative binomial distribution is Gamma-Poisson mixture distribution, and now we know why: the dictating λ comes from a continuous mixture. That’s what we needed to explain the heterogeneity amongst sellers.

Let’s simulate this scenario to gain more intuition.

First, we draw λᵢ from a Gamma distribution: λᵢ ~ Γ(r, θ). Intuitively, the Gamma distribution tells us about the variety in the intensity — listing rate — amongst the sellers.

On a practical note, one can instill their assumptions about the degree of heterogeneity in this step of the model: how different are sellers? By varying the levels of heterogeneity, one can observe the impact on the final Poisson-like distribution. Doing this type of checks (i.e., posterior predictive check), is common in Bayesian modeling, where the assumptions are set explicitly.

*Gamma-Poisson mixture distribution versus homogenous Poisson distribution. Τhe dashed line reflects λ, which is 4 for both distributions.*

In the second step, we plug the obtained λ into the Poisson distribution: 𝑋 | λ ~ Poisson(λ), and obtain a Poisson-like distribution that represents the summed subprocesses. Notably, this unified process has a larger dispersion than expected from a homogeneous Poisson distribution, but it is in line with the Gamma mixture of λ.

Heterogeneous λ and inference

A practical consequence of introducing flexibility into your assumed distribution is that inference becomes more challenging. More parameters (i.e., the Gamma parameters) need to be estimated. Parameters act as flexible explainers of the data, tending to overfit and explain away variance in your variable. The more parameters you have, the better the explanation may seem, but the model also becomes more susceptible to noise in the data. Higher variance reduces the power to identify a difference in means, if one exists, because — well — it gets lost in the variance.

Countering the loss of power

Confirm whether you indeed need to extend the standard Poisson distribution. If not, simplify to the simplest, most fit model. A quick check on overdispersion may suffice for this.
Pin down the estimates of the Gamma mixture distribution parameters using regularising, informative priors (think: Bayes).

During my research process for writing this blog, I learned a great deal about the connective tissue underlying all of this: how the binomial distribution plays a fundamental role in the processes we’ve discussed. And while I’d love to ramble on about this, I’ll save it for another post, perhaps. In the meantime, feel free to share your understanding in the comments section below 👍.

Conclusion

The Poisson distribution is a simple distribution that can be highly suitable for modelling count data. However, when the assumptions do not hold, one can extend the distribution by allowing the rate parameter to vary as a function of time or other factors, or by assuming subprocesses that collectively make up the count data. This added flexibility can address the limitations, but it comes at a cost: increased flexibility in your modelling raises the variance and, consequently, undermines the statistical power of your model.

If your end goal is inference, you may want to think twice and consider exploring simpler models for the data. Alternatively, switch to the Bayesian paradigm and leverage its built-in solution to regularise estimates: informative priors.

I hope this has given you what you came for — a better intuition about the Poisson distribution. I’d love to hear your thoughts about this in the comments!

Unless otherwise noted, all images are by the author.
Originally published at https://aalvarezperez.github.io on January 5, 2025.

Stay Ahead

Explore More Insights

Stay ahead with more perspectives on cutting-edge power, infrastructure, energy, bitcoin and AI solutions. Explore these articles to uncover strategies and insights shaping the future of industries.

Cornelis lands $205M to make AI networks compute, not just connect

In Ethernet mode, the CN6000 can pair with switches from other vendors, including Broadcom’s Tomahawk line. Traffic sent over Ethernet still passes through an internal encapsulation layer built around Omni-Path. Spelman said that internal architecture preserves congestion management, credit-based flow control, and packet spraying, aimed at low latency and high

Arm expands AI chip push with Neoverse CSS N4 and AGI CPU

Arm claimed that the processor could provide more than twice the performance per rack of comparable x86 systems, although that figure is based on Arm’s internal estimates and depends on system configuration and workloads. The AGI CPU contains up to 136 Neoverse V3 cores and has a 300 watt TDP.

AI inferencing is headed for the network edge

On the chip front, AI edge models run on an advanced type of chip called a neural processing unit or NPU. These chips, such as Google’s Tensor Processing Unit (TPU) or Qualcomm’s Snapdragon, are small, efficient, and don’t use as much energy or create as much heat as traditional CPUs

Broadcom hampers VMware migration by blocking downloads of key SDK

Broadcom is raising the stakes when it comes to preventing its VMware customers from moving to another platform. The company has stopped users from downloading its Virtual Disk Development Kit (VDDK). While this may not look like significant news VDDK is, in fact, a vital part of tools used to

Insights: Prioritizing process safety management over profits across the refining industry (Pt. 1)

In this Insights episode of the Oil & Gas Journal ReEnterprised podcast, downstream editor and lead reporter Robert Brelsford talks with Joe Barnes, principal of Barnes’ Engineering Consultants and a veteran oil and gas operations, maintenance, reliability, and major-projects leader with more than 30 years of industry experience. Barnes, who spent roughly half of his career in refining and half in exploration and production, discusses why process safety management must remain a central concern for everyone working in oil and gas—not only safety professionals. Drawing on his experience with a major refinery accident and the aftermath of a separate major offshore incident, Barnes explains why good intentions are not enough to prevent major incidents, particularly in times when the potential for high margins could tempt operator’s into promoting maximized production over execution of maintenance on processing and production installations to ensure safe and reliable operations. During the conversation, Robert and Joe examine recurring root causes identified through Barnes’ review of process-safety incidents dating back to the mid-1980s, including absence of leadership, poor communication, inadequate risk identification and mitigation, aging equipment, deferred maintenance, workforce turnover, and the loss of experienced personnel. They also discuss why the industry has sometimes failed to consistently share and apply lessons from previous incidents, particularly when investigation findings are not widely accessible. The two consider the difference between personal safety and process safety metrics, the warning signs that preceded the 2005 BP Texas City refinery explosion, and the role of management of change (MOC) in evaluating staffing, maintenance, turnaround, efficiency, and capital-reduction decisions. Barnes also outlines the questions boards and executive leaders should be asking to ensure that process safety remains a business priority across economic cycles. Robert and Joe further explore how refineries can use past incident scenarios as practical training tools, strengthen external oversight

Energy Department Launches $16 Million Prize To Grow the Mining and Critical Minerals Workforce

WASHINGTON—The U.S. Department of Energy’s (DOE) Office of Critical Minerals and Energy Innovation (CMEI) today launched the PROSPECT Planning Prize to support universities in developing the skilled workforce required to produce, process, recover, and recycle critical minerals on American soil. Each winner will receive up to $1 million from a total prize pool of $16 million. “Mining and critical minerals professionals are essential to America’s success in manufacturing, AI, energy, and national security—but this country isn’t producing enough of them,” said Assistant Secretary of Energy Audrey Robertson. “Through this prize competition, DOE will empower universities to pursue more ambitious strategies for attracting and educating the workers we need for a robust critical mineral supply chain.” This prize competition is part of DOE’s $100 million Providing Opportunities for Specialized Education in Critical Technologies (PROSPECT) initiative, which delivers on President Trump’s bold agenda to restore American energy dominance, rebuild domestic supply chains, and reduce reliance on foreign adversaries for critical minerals. The initiative’s near-term goal is to double the number of graduates with mining, minerals, and associated supply chain credentials across the United States. U.S.-based academic institutions with prize-eligible degree programs are encouraged to submit plans. Proposed programs should aim to substantially increase the number of graduates and other credentialed professionals prepared for relevant mining and supply chain careers in the next two years. Programs may target a range of beneficiaries, including early-career professionals, incumbent workers, educators, and students in university, community college, or high school. Applications are now open. The deadline for submission is Oct. 5, 2026. Learn more about this competition, including key dates and submission details, on the PROSPECT Prize page. The PROSPECT Planning Prize is funded by CMEI and managed by the National Laboratory of the Rockies under the American-Made prize program. ###

U.S. Secretary of Energy Chris Wright Delivers U.S. National Statement at the General Conference of the International Atomic Energy Agency in Vienna, Austria

VIENNA, AUSTRIA—U.S. Secretary of Energy Chris Wright today delivered the U.S. National Statement at the General Conference of the International Atomic Energy Agency (IAEA) in Vienna, Austria. Secretary Wright’s full remarks from the International Atomic Energy Agency (IAEA) General Conference are below: I am honored to represent the United States of America at the 70th IAEA General Conference. This year, as we celebrate the 250th anniversary of the founding of the United States, we reflect on a history defined by a tireless spirit of innovation and bold exploration. It was this American drive that, over seven decades ago, moved President Dwight D. Eisenhower to deliver his historic ‘Atoms for Peace’ address, sowing the seeds for the creation of this very Agency. Together, our high-level cooperation has achieved monumental milestones—pioneering global safeguards, securing vulnerable fissile materials across the globe, countering and preparing for radiological and nuclear threats, creating energy abundance, and advancing nuclear science to improve living standards. As the world’s foremost nuclear innovator, the United States remains steadfastly dedicated to the IAEA’s mission. I would like to take a moment to congratulate Timor Leste on becoming the newest member of the IAEA. America looks forward to working with you. Thanks to President Trump’s leadership, the United States is ushering in an American nuclear renaissance. President Trump directed the Administration to revitalize America’s nuclear sector, modernizing regulations, improving processes and streamlining advanced reactor testing to rebuild the nuclear industrial base. This agenda allows the United States to establish an expedited pathway for testing and approving advanced reactors, set standards for converting surplus uranium and plutonium into reactor fuel, and support the energy infrastructure needed for America’s reindustrialization and artificial intelligence. These actions advance President Trump’s bold goal of adding 300 gigawatts of new nuclear capacity in the United States by 2050.

YPF taps Axens for new diesel hydrotreater at Argentinian refinery

YPF SA has let a contract to Axens Group to license proprietary technology and provide associated equipment for a new 5,000-cu m/day diesel hydrotreating unit to be installed at one of the Argentine operator’s in-country industrial complexes. As part of the contract announced in September, Axens will deliver the new diesel hydrotreating unit—which will be configured with Prime-D technology—in prefabricated modules, an approach the technology provider said will reduce project execution time by 6 months compared with conventional site construction. The modular design will integrate pressure vessels, structural components, piping, and instrumentation before shipment to the refinery. The approach is intended to limit on-site construction and interference with ongoing refinery operations, Axens said. While Axens confirmed it will execute the project in a way that allows YPF to maintain ongoing operations of the refinery, neither the service provider nor YPF identified which industrial complex will receive the new unit or disclose its capital cost, schedule, or anticipated startup date. Broader fuel-improvement program The new unit comes as part of YPF’s broader program to bring its refining system into compliance with Argentina’s tighter diesel specifications. YPF operates three refineries with combined capacity of 338,000 b/d: La Plata, Luján de Cuyo, and Plaza Huincul. In the operator’s second-quarter 2026 earnings report and accompanying presentation, YPF confirmed completing installation and commissioning in July of its previously announced new hydrotreating unit at the 113,900-b/d Luján de Cuyo refining complex in Mendoza Province, Argentina. In its 2025 annual report published earlier in 2026, YPF said the Luján de Cuyo project also was to include a new hydrogen-generation unit, as well as a revamp of a separate hydrotreating unit at the site. The report identified an estimated $637 million investment for the refinery’s wider fuel-quality program. The annual report also confirmed advanced engineering design for new

IEA: Ukrainian drone campaign degrades Russian refining resilience

Increasingly frequent and precise Ukrainian drone attacks are damaging Russian processing units and lengthening repair times, forcing Moscow into unprecedented export bans, relaxed fuel-quality standards, and imports to protect domestic supply, according to the International Energy Agency (IEA). Russia’s 32 major refineries, with 6.5 million b/d of nameplate capacity, make it the world’s third-largest refined-products producer after the US and China. But crude runs fell to 3.8 million b/d in June 2026, down 30% year on year and the lowest since May 2004. Gasoline output is reportedly down about 20%. Ukraine has targeted Russian oil infrastructure since the 2022 full-scale invasion, but IEA said the campaign’s scale, range, and effectiveness increased sharply during 2025-26. A Russian refinery was successfully struck once every 3 days on average in the first 8 months of 2026. Ukrainian forces are now sending multiple drone waves against single sites, overwhelming protective netting and air defenses. Reach has expanded as well. On July 7, Ukraine struck Gazprom Neft PJSC’s 450,000-b/d Omsk refinery, Russia’s largest, some 2,500 km from the border, while several refineries closer to Ukraine have been hit as many as 15 times. By late August, only four major refineries—all in Eastern Siberia or farther east, 3,500-6,500 km from Ukraine—remained untouched. Secondary units targeted Targeting also appears more precise, increasingly hitting secondary units alongside crude distillation units. Fluid catalytic crackers, reformers, and hydrotreaters are critical to light-product yields and fuel specifications. Minor damage to a crude distillation unit can often be repaired in 1-2 weeks, but serious damage to more complex units can require 6-8 months, IEA said. The 250,000-b/d Moscow refinery, heavily damaged in June, is reportedly offline until early 2027. Sanctions are compounding the problem by restricting access to replacement equipment and specialist suppliers. Refiners have tried to preserve throughput by postponing maintenance,

BLM Wyoming lease sale generates $82 million, Colorado’s sale fetches $4.7 million

The US Bureau of Land Management (BLM)’s quarterly lease sale in Wyoming Sept. 10 showed a robust $82.4 million in high bids, while its Colorado lease sale pulled in $4.7 million, a fraction of the $35.26 million in revenue generated at the bureau’s last lease sale in the state in June. In Wyoming, BLM leased 99 parcels totaling 114,389 acres, with strong interest and competition in Converse and Campbell Counties in the Powder River basin of northeastern Wyoming. The two counties produced 64% of Wyoming’s crude oil in 2024, with Converse leading at 43.3 million bbl, according to the Wyoming State Geological Survey. BLM’s sales statistics show that of the 17 parcels receiving 10 or more bids during the sale, all but 2 were in Converse and Campbell Counties. Most parcels there fetched over $1000/acre, with one in Converse receiving an $8,000/acre winning bid. In contrast, many other counties saw little competition, with 37,000 acres receiving no bids and 12 parcels, totaling over 16,000 acres, going for the legal minimum of $10/acre. In Colorado, BLM leased 29 parcels totaling 14,212 acres. The bureau offered far more acreage in the two previous sales–leasing 134,173 acres in June and 42,532 acres in September. Details of the Colorado sale were unavailable pending the state’s final review. Federal onshore oil and gas leases extend 10 years or as long as production continues in paying quantities, and they carry a 12.5% royalty rate, with revenues split between the federal government and the state.

Nuclear’s Next AI Test: Building at Scale

For data center developers facing multiyear utility interconnection queues and tightening power markets, nuclear energy is entering a different phase of its AI infrastructure story. The near-term opportunity still rests largely with the existing reactor fleet. Holtec International has moved the Palisades Nuclear Plant in Michigan into fuel loading, one of the final major stages before reactor startup activities. Constellation Energy, meanwhile, continues to work toward a 2027 restart of the former Three Mile Island Unit 1, now the Christopher M. Crane Clean Energy Center, under its long-term power agreement with Microsoft. Together, Palisades and Crane represent roughly 1.64 GW of existing nuclear capacity that could return to service without waiting for entirely new plants to be licensed, financed and constructed. That makes reactor restarts one of the few ways nuclear generation can materially intersect with data center power demand before the end of the decade. But the more consequential change may be taking place further upstream. A burst of activity from advanced nuclear developers at the end of August pointed increasingly toward the industrial systems required to move new reactor designs from demonstrations to repeatable infrastructure. X-energy, TerraPower, GE Vernova Hitachi, Oklo, Westinghouse, Kairos Power and others reported progress involving fuel supply, reactor testing, manufacturing, licensing and commercial deployment. None of these advanced reactor projects will solve the industry’s 2027 or 2028 power shortage. That is no longer the most useful test. The more important question is whether advanced nuclear can begin acquiring the characteristics of an industrial supply chain: dependable fuel, standardized manufacturing, repeatable construction, tested reactor systems and enough commercial certainty for large power customers to plan around deployment schedules measured in years rather than speculation. For data center infrastructure, that is the transition worth watching. Palisades Moves From Restoration to Startup The clearest near-term proof point

The waste heat paradox: How data centers can cool AI with the heat AI creates

A 40-year-old technology built for this exact moment Absorption chillers produce chilled water the way a conventional chiller does, but they use heat instead of electricity to drive the refrigeration cycle. A generator uses hot water, steam, or exhaust gas to separate refrigerant vapor from a lithium bromide solution; the vapor condenses, evaporates under low pressure to produce the cooling effect, and is reabsorbed to close the loop. Because there’s no large electrically driven compressor, the electrical footprint is small: industry analysis puts absorption chillers at roughly 2 MW of cooling output for just 20–25 kW of electrical input, compared to 500 kW or more of electrical draw for a conventional chiller doing the same job. The technology has existed commercially for decades and never displaced electric chillers at scale, for one simple reason: it needs a steady, moderate-to-high-temperature heat source to run, and building a boiler specifically to feed one erased most of the savings. That missing piece is exactly what two current infrastructure trends are now supplying as a byproduct. Two trends CIOs are already funding that solve the “missing heat” problem 1. On-site power generation is becoming standard, not exceptional Grid interconnection delays in major data center markets now stretch into years, pushing hyperscale operators toward gas turbines, engines, and fuel cells built directly on campus — all of which reject substantial heat as a byproduct of making electricity. Bloom Energy already pairs its fuel-cell systems directly with absorption chillers at data center sites, using exhaust heat to generate chilled water and reduce reliance on the electric chiller plant. Some industry forecasts expect roughly a third of data centers to run fully on-site-powered campuses by 2030 — meaning this heat stream is a permanent feature of the infrastructure roadmap, not a one-off opportunity.

DCF Poll: How Should Utilities Vet Real vs. ‘Ghost’ Data Center Demand?

Matt Vincent is Editor in Chief of Data Center Frontier, where he leads editorial strategy and coverage focused on the infrastructure powering cloud computing, artificial intelligence, and the digital economy. A veteran B2B technology journalist with more than two decades of experience, Vincent specializes in the intersection of data centers, power, cooling, and emerging AI-era infrastructure. Since assuming the EIC role in 2023, he has helped guide Data Center Frontier’s coverage of the industry’s transition into the gigawatt-scale AI era, with a focus on hyperscale development, behind-the-meter power strategies, liquid cooling architectures, and the evolving energy demands of high-density compute, while working closely with the Digital Infrastructure Group at Endeavor Business Media to expand the brand’s analytical and multimedia footprint. Vincent also hosts The Data Center Frontier Show podcast, where he interviews industry leaders across hyperscale, colocation, utilities, and the data center supply chain to examine the technologies and business models reshaping digital infrastructure. Since its inception he serves as Head of Content for the Data Center Frontier Trends Summit. Before becoming Editor in Chief, he served in multiple senior editorial roles across Endeavor Business Media’s digital infrastructure portfolio, with coverage spanning data centers and hyperscale infrastructure, structured cabling and networking, telecom and datacom, IP physical security, and wireless and Pro AV markets. He began his career in 2005 within PennWell’s Advanced Technology Division and later held senior editorial positions supporting brands such as Cabling Installation & Maintenance, Lightwave Online, Broadband Technology Report, and Smart Buildings Technology. Vincent is a frequent moderator, interviewer, and keynote speaker at industry events including the HPC Forum, where he delivers forward-looking analysis on how AI and high-performance computing are reshaping digital infrastructure. He graduated with honors from Indiana University Bloomington with a B.A. in English Literature and Creative Writing and lives in southern New Hampshire with

Nvidia invests $3.5B in MediaTek to extend its grip on AI

AI infrastructure: MediaTek will work with Nvidia’s NVLink Fusion ecosystem to enable customers to develop custom AI infrastructure designed to integrate with Nvidia rack-scale systems and AI factories. Local AI computing: The companies will continue to collaborate on multiple generations of Nvidia RTX Spark and DGX Spark PC chips, powering consumer PCs, AI developer supercomputers, and enterprise-class workstations, that integrate Nvidia GPUs with MediaTek SoCs. Automotive: MediaTek and Nvidia will continue developing platforms for AI-powered, software-defined vehicles in the era of physical AI. “MediaTek is one of the world’s great semiconductor companies, with exceptional expertise in system-on-chip design, connectivity, leading performance and power efficiency,” said Jensen Huang, founder and CEO of Nvidia, in a statement. “Together, we’re building platforms that bring Nvidia accelerated computing to new markets and give customers the freedom to create differentiated AI systems at enormous scale.” NVLink is a high-performance interface, but this move also helps lock in customers to the Nvidia platform, since NVLink Fusion is not about to hook up to AMD processors. For MediaTek, the Nvidia investment provides a big pile of cash and access to Nvidia’s infrastructure as it attempts to establish itself as a major supplier of custom data center silicon.

AI Puts Fiber on the Critical Path

Fiber Becomes Part of the Build A decade ago, DC Blox might have prioritized land near existing fiber from AT&T, Verizon, Zayo or another established network provider. That calculation is different today. As data center campuses have grown and hyperscalers have become a larger share of the customer base, Wabik said connectivity has increasingly become another construction package associated with the project itself. “If Zayo or Verizon or AT&T just happens to be there close by, that’s a good thing,” he said. “But fiber construction is inherently anymore just part of the construction component.” That changes the site-selection question from Is there fiber nearby? to Can fiber be built here at the scale and diversity the customer requires? For DC Blox, Wabik said that can mean assessing whether sufficient public right-of-way exists to establish three or sometimes four diverse fiber paths into a data center. That distinction is important, as AI workloads push infrastructure into markets where power, land and energy options may be more abundant than established carrier density. The hyperscalers themselves have also become major network builders. Wabik characterized them provocatively as today’s telecom providers, pointing to the scale of terrestrial fiber they commission as well as the growing role of companies such as Amazon, Google and Meta in subsea cable development. The point is less that traditional carriers have disappeared than that hyperscalers increasingly design, commission and control enormous portions of the connectivity required to support their own infrastructure. DC Blox now sees requests for 864-count fiber as routine and, in some cases, 1,728-count cable. That would have been difficult to imagine during an earlier era when a handful of fibers from an established carrier could satisfy a data center’s connectivity requirements. AI-Scale Fiber Gets Physical The scale becomes clearer when the discussion moves from abstract network

Power First: AI Data Centers Become Energy Systems

For decades, data centers consumed electricity much like other large commercial customers: power arrived from the utility, while batteries and diesel generators stood behind it to protect the load. AI is starting to break that model. As data center campuses grow toward hundreds of megawatts and, in some cases, gigawatt scale, developers are increasingly taking responsibility for an energy system that once sat largely outside the data center boundary. Natural gas supply, onsite generation, fuel cells, batteries, controls and the behavior of the compute load itself are increasingly becoming parts of the same infrastructure system. That was the central thread running through “Power First: The New Playbook for Delivering AI Data Centers,” an Aug. 4 session at the Data Center Frontier Trends Summit 2026 in Reston, Virginia. Moderated by Fengrong Li, Senior Managing Director at FTI Consulting, the panel brought together Jim Summers, CEO of GPC Infrastructure; Shankar Achanta, EVP and Chief Product and Technology Officer at FuelCell Energy; Judith Judson, Executive Vice President at Calibrant Energy; and Yuval Bachar, Founder and CEO of EdgeCloudLink. The discussion began with the immediate constraint — the grid cannot deliver capacity on the timetable AI developers increasingly require — but quickly moved beyond the familiar concept of “bridge power.” The larger question was what happens when the data center itself becomes an energy system. From Backup Power to Prime Power Behind-the-meter generation is not new. What has changed is its role and scale. “Traditionally, behind-the-meter generation has been for backup and the sizes were smaller,” Achanta said. “But what they’re seeing is the demand for the power is growing rapidly due to the data center load.” Interconnection queues, transmission limitations and equipment supply constraints are pushing onsite generation into what Achanta called the “front seat,” supplying primary power rather than waiting behind the

Microsoft will invest $80B in AI data centers in fiscal 2025

And Microsoft isn’t the only one that is ramping up its investments into AI-enabled data centers. Rival cloud service providers are all investing in either upgrading or opening new data centers to capture a larger chunk of business from developers and users of large language models (LLMs). In a report published in October 2024, Bloomberg Intelligence estimated that demand for generative AI would push Microsoft, AWS, Google, Oracle, Meta, and Apple would between them devote $200 billion to capex in 2025, up from $110 billion in 2023. Microsoft is one of the biggest spenders, followed closely by Google and AWS, Bloomberg Intelligence said. Its estimate of Microsoft’s capital spending on AI, at $62.4 billion for calendar 2025, is lower than Smith’s claim that the company will invest $80 billion in the fiscal year to June 30, 2025. Both figures, though, are way higher than Microsoft’s 2020 capital expenditure of “just” $17.6 billion. The majority of the increased spending is tied to cloud services and the expansion of AI infrastructure needed to provide compute capacity for OpenAI workloads. Separately, last October Amazon CEO Andy Jassy said his company planned total capex spend of $75 billion in 2024 and even more in 2025, with much of it going to AWS, its cloud computing division.

John Deere unveils more autonomous farm machines to address skill labor shortage

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Self-driving tractors might be the path to self-driving cars. John Deere has revealed a new line of autonomous machines and tech across agriculture, construction and commercial landscaping. The Moline, Illinois-based John Deere has been in business for 187 years, yet it’s been a regular as a non-tech company showing off technology at the big tech trade show in Las Vegas and is back at CES 2025 with more autonomous tractors and other vehicles. This is not something we usually cover, but John Deere has a lot of data that is interesting in the big picture of tech. The message from the company is that there aren’t enough skilled farm laborers to do the work that its customers need. It’s been a challenge for most of the last two decades, said Jahmy Hindman, CTO at John Deere, in a briefing. Much of the tech will come this fall and after that. He noted that the average farmer in the U.S. is over 58 and works 12 to 18 hours a day to grow food for us. And he said the American Farm Bureau Federation estimates there are roughly 2.4 million farm jobs that need to be filled annually; and the agricultural work force continues to shrink. (This is my hint to the anti-immigration crowd). John Deere’s autonomous 9RX Tractor. Farmers can oversee it using an app. While each of these industries experiences their own set of challenges, a commonality across all is skilled labor availability. In construction, about 80% percent of contractors struggle to find skilled labor. And in commercial landscaping, 86% of landscaping business owners can’t find labor to fill open positions, he said. “They have to figure out how to do

2025 playbook for enterprise AI success, from agents to evals

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More 2025 is poised to be a pivotal year for enterprise AI. The past year has seen rapid innovation, and this year will see the same. This has made it more critical than ever to revisit your AI strategy to stay competitive and create value for your customers. From scaling AI agents to optimizing costs, here are the five critical areas enterprises should prioritize for their AI strategy this year. 1. Agents: the next generation of automation AI agents are no longer theoretical. In 2025, they’re indispensable tools for enterprises looking to streamline operations and enhance customer interactions. Unlike traditional software, agents powered by large language models (LLMs) can make nuanced decisions, navigate complex multi-step tasks, and integrate seamlessly with tools and APIs. At the start of 2024, agents were not ready for prime time, making frustrating mistakes like hallucinating URLs. They started getting better as frontier large language models themselves improved. “Let me put it this way,” said Sam Witteveen, cofounder of Red Dragon, a company that develops agents for companies, and that recently reviewed the 48 agents it built last year. “Interestingly, the ones that we built at the start of the year, a lot of those worked way better at the end of the year just because the models got better.” Witteveen shared this in the video podcast we filmed to discuss these five big trends in detail. Models are getting better and hallucinating less, and they’re also being trained to do agentic tasks. Another feature that the model providers are researching is a way to use the LLM as a judge, and as models get cheaper (something we’ll cover below), companies can use three or more models to

OpenAI’s red teaming innovations define new essentials for security leaders in the AI era

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More OpenAI has taken a more aggressive approach to red teaming than its AI competitors, demonstrating its security teams’ advanced capabilities in two areas: multi-step reinforcement and external red teaming. OpenAI recently released two papers that set a new competitive standard for improving the quality, reliability and safety of AI models in these two techniques and more. The first paper, “OpenAI’s Approach to External Red Teaming for AI Models and Systems,” reports that specialized teams outside the company have proven effective in uncovering vulnerabilities that might otherwise have made it into a released model because in-house testing techniques may have missed them. In the second paper, “Diverse and Effective Red Teaming with Auto-Generated Rewards and Multi-Step Reinforcement Learning,” OpenAI introduces an automated framework that relies on iterative reinforcement learning to generate a broad spectrum of novel, wide-ranging attacks. Going all-in on red teaming pays practical, competitive dividends It’s encouraging to see competitive intensity in red teaming growing among AI companies. When Anthropic released its AI red team guidelines in June of last year, it joined AI providers including Google, Microsoft, Nvidia, OpenAI, and even the U.S.’s National Institute of Standards and Technology (NIST), which all had released red teaming frameworks. Investing heavily in red teaming yields tangible benefits for security leaders in any organization. OpenAI’s paper on external red teaming provides a detailed analysis of how the company strives to create specialized external teams that include cybersecurity and subject matter experts. The goal is to see if knowledgeable external teams can defeat models’ security perimeters and find gaps in their security, biases and controls that prompt-based testing couldn’t find. What makes OpenAI’s recent papers noteworthy is how well they define using human-in-the-middle