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.

Noble Quarterly Revenue Falls

Noble Corp on Monday reported $798 million in revenue for the third quarter, down from $849 million for the prior three-month period as lower rig utilization offset lower contract drilling services costs. “Utilization of the 35 marketed rigs was 65 percent in the third quarter of 2025 compared to 73

Google Cloud targets enterprise AI builders with upgraded Vertex AI Training

Enterprises can quickly set up managed Slurm environments with automated resiliency and cost optimization through the Dynamic Workload Scheduler. The platform also includes hyperparameter tuning, data optimization, and built-in recipes with frameworks like NVIDIA NeMo to streamline model development. Enterprises weigh AI training gains Building and scaling generative AI models

Quantum Circuits brings dual-rail qubits to Nvidia’s CUDA-Q development platform

Nvidia announced AWS Braket support at the end of 2024. CUDA-Q is also available on Nvidia’s own Quantum Cloud platform, which combines GPUs and quantum processors with AI. The other major quantum computing platform is IBM’s Qiskit, which Quantum Circuits was already supporting. Qiskit also works with IBM, IonQ, Rigetti,

Japan’s semiconductor industry battles back

Sekiguchi Waichi, CEO, MM Research Institute, Inc.

XRG in Talks to Invest in Argentina LNG Project

The overseas unit of Abu Dhabi’s biggest oil company is in talks to invest in a liquefied natural gas project Argentina’s YPF SA is developing as it pushes to start exporting the fuel, according to people familiar with the matter. XRG is eyeing a stake in the project as it considers expanding its LNG portfolio in Latin America, the US and Asia, according to one of the people, who asked not to be identified because the matter isn’t public. State-run YPF is developing the floating terminal as Argentina tries to tap global LNG demand and accelerate output of vast natural gas reserves in the Vaca Muerta shale basin. The project, which requires construction of several liquefaction vessels, is designed to eventually produce 28 million tons of LNG annually. Shell Plc and Eni SpA are working with YPF on the project, but final investment decisions haven’t been made. XRG’s talks with YPF are preliminary, and the company may ultimately decide not to pursue an investment, according to the people. XRG declined to comment. On Monday, YPF’s depositary receipts traded in New York jumped as much as 38% after libertarian President Javier Milei’s party prevailed in legislative elections. The midterms were seen as a pivotal moment for foreign investors looking for opportunities in Argentina, providing a clear sign voters would continue backing Milei’s push to deregulate the economy. XRG is the international arm of Abu Dhabi National Oil Co., backed by Abu Dhabi’s oil wealth. It has already acquired a stake in NextDecade Corp.’s Rio Grande LNG project being built in South Texas, and is in the process of taking over Germany’s Covestro AG as it bets on lasting demand for gas and chemicals in the energy transition. It’s also bought gas assets in Africa and Central Asia. But the company also had a setback last month when it

Venezuela Revokes Trinidad Gas Deals Over USA Alliance

Venezuela revoked energy deals with neighboring Trinidad and Tobago for its support of a US military offensive in the Caribbean, potentially raising the economic cost of the twin-island nation’s alliance with the Trump administration. Speaking on state television Monday evening, President Nicolás Maduro revoked an energy framework agreement with Trinidad that allowed the two countries to forge gas deals. Venezuelan Vice President and Oil Minister Delcy Rodríguez had made the proposal earlier on Monday. “Faced with the Prime Minister’s threat to turn Trinidad into the aircraft carrier of the US empire against Venezuela, against South America, there is only one alternative,” Maduro said in his weekly program on state TV. “It is completely suspended.” Maduro said he would deliver the proposal to the Supreme Court, National Assembly and State Council to receive their recommendations before taking “a structural measure” very soon. Trinidad needs Venezuelan gas to replenish supply to the industrial backbone of its fragile economy. However, when Rodríguez first made the threat on Monday, Trinidad’s Prime Minister Kamla Persad-Bissessar told AFP that the country’s future “does not depend on Venezuela and never has.” Trinidad’s government has developed a “hostile attitude” and a “warlike plan” against Venezuela, Rodríguez said during the afternoon, by “siding” with the US’s “military agenda.” Persad-Bissessar has previously said she welcomed the US offensive on drug traffickers, calling for them to be killed “violently.” Her posture has alienated Trinidad from other English-language countries in the Caribbean that have insisted on maintaining the region as a “zone of peace.” Venezuelan rhetoric is escalating as the US advances a military campaign, blowing up purported drug boats and pointing a finger at Maduro and, increasingly, Colombian President Gustavo Petro for allegedly flooding the US with fentanyl and cocaine. The US is ramping up its deployment in the southern Caribbean, with a

Energy CEOs ‘Optimistic About Growth’

In a statement sent to Rigzone on Monday, KPMG noted that, according to its 2025 Global Energy, Natural Resources, and Chemicals CEO Outlook, “CEOs in the energy, natural resources, and chemicals (ENRC) sector are optimistic about growth”. “Despite inflation and regulatory headwinds, confidence among CEOs is rising, with 84 percent optimistic about mid-term industry growth – up from 72 percent in 2024,” the statement said. “This outlook is driven by strong demand for both fossil fuels and renewables, alongside innovation in energy storage, smart grids, and carbon capture,” it added. “While 78 percent remain positive about their own company’s growth, a slight dip from last year reflects concerns over shifting regulations, trade volatility, and inflationary pressures – particularly in the chemicals sector,” it continued. The statement went on to note that M&A strategies “are also evolving with just a few CEOs (36 percent) expecting to pursue ‘high impact’ deals in 2025, down from 58 percent in 2024, while 55 percent anticipate ‘moderate’ deal activity, a rise from 38 percent last year, suggesting a shift toward more cautious growth strategies”. The KPMG statement highlighted that artificial intelligence has “rapidly evolved into a core strategy in the energy sector”. It pointed out that 65 percent of CEOs now rank generative AI as a top investment, which it noted is up 12 points from 2024, and that 72 percent are planning to allocate 10-20 percent of their budgets to AI over the next year. “ROI expectations are climbing, with 66 percent anticipating returns within one to three years, up significantly in comparison to just 15 percent last year,” KPMG said in the statement. Momentum is also building around agentic AI, according to the statement, which highlighted that 51 percent of CEOs expect it to transform operations and workforce efficiency. “Yet despite growing confidence,

Google-NextEra, Santee Cooper announcements signal new life for defunct nuclear projects

Dive Brief: Google and NextEra Energy will collaborate to restart the 600-MW Duane Arnold Energy Center in Iowa as part of a larger partnership aimed at accelerating nuclear deployment across the U.S., the two companies said Monday. On Friday, Santee Cooper revealed it has signed a letter of intent regarding the potential sale of two unfinished reactors at the abandoned V.C. Summer project in South Carolina to Brookfield Asset Management. The 2.2-GW project was mothballed in 2017 following delays and cost overruns. The reexamination of retired or canceled nuclear projects comes as the U.S. is desperate for power resources to support artificial intelligence and other demand centers. New resources may have some market advantages over existing generators, and the recent deals are a “positive” for the sector, according to equity analysts at Jefferies. Dive Insight: The U.S. Department of Energy last week directed the Federal Energy Regulatory Commission to initiate rulemaking to accelerate large loads interconnecting to the nation’s electric grid. While the DOE’s action is broadly seen as designed to speed the interconnection of AI and data center loads, Jefferies analyst Julien Dumoulin-Smith said the market may have overlooked the proposed rule’s recommendation that existing generators co-locating with new loads 20 MW or higher be required to undertake a reliability review. “The generator could not serve new load until necessary transmission upgrades are completed, at the generators expense,” Dumoulin-Smith wrote in a Monday note about the DOE letter and new nuclear deals. “We see the letter as all but ensuring the market will prefer front-of-the meter deployments, with preferential treatment clearly emphasized for those bringing new supply.” In South Carolina, Jefferies sees Brookfield going “all-in on new nuclear.” It said “Brookfield’s participation in nuclear is bullish for new nuclear companies broadly.” Proposed in 2008, the V.C. Summer expansion was supposed

US partners with Westinghouse, Cameco and Brookfield on $80B nuclear deployment

Westinghouse Electric, Cameco and Brookfield Asset Management have entered into a strategic partnership with the U.S. government to deploy $80 billion in new nuclear reactors, the companies announced Tuesday. “Our administration is focused on ensuring the rapid development, deployment, and use of advanced nuclear technologies. This historic partnership supports our national security objectives and enhances our critical infrastructure,” Secretary of Commerce Howard Lutnick said in a statement. The deal calls for Westinghouse AP1000 reactors to be utilized in a deployment that will create more than 100,000 construction jobs, the companies said. “The program will cement the United States as one of the world’s nuclear energy powerhouses and increase exports of Westinghouse’s nuclear power generation technology globally.” According to the announcement, the partnership contains “profit sharing mechanisms” that allow all parties, “including the American people,” to participate in the “long-term financial and strategic value that will be created within Westinghouse by the growth of nuclear energy and advancement of investment into AI capabilities in the United States.” Brookfield has more than half a trillion dollars invested in the critical infrastructure that underpins the U.S. economy, “and we expect to double that investment in the next decade as we deliver on building the infrastructure backbone of artificial intelligence,” Brookfield President Connor Teskey said in a statement. The U.S. is trying to rapidly bring power resources online to meet rising demand from data centers, and interest in nuclear is growing. Following a decade of stagnant growth, U.S. electricity demand will increase at a 2.5% compound annual growth rate through 2035, according to Bank of America Institute research published in July. “For all of the energy policy disagreements in Washington, one thing is clear: nuclear energy is the baseload electrical power source of the future,” said Thomas Ryan, the managing partner of K&L Gates’

Kentucky utility, AG challenge PJM’s regional cost allocation plan for DOE ‘emergency’ orders

Dive Brief: East Kentucky Power Cooperative and Kentucky’s attorney general are challenging the PJM Interconnection’s plan to make utilities and other load-serving entities across its footprint pay costs related to power plants that run under U.S. Department of Energy emergency orders. Utilities that aren’t responsible for the resource adequacy concerns driving the DOE’s emergency orders under the Federal Power Act’s section 202(c) should not have to pay for those costs, EKPC and the attorney general said in a joint protest filed on Monday at the Federal Energy Regulatory Commission. “LSEs who actively manage the resource adequacy needs of the load they are obligated to serve should not be penalized by being forced to pay socialized costs that are driven by the other LSEs’ lack of active resource management,” EKPC and the attorney general said. Dive Insight: In PJM, DOE has issued orders directing Constellation Energy to operate two 380-MW gas- and oil-fired units at its Eddystone power plant near Philadelphia past their planned retirement date. On Friday, DOE issued an emergency order at PJM’s request to allow a nearly 400-MW oil-fired unit near Baltimore to run beyond its operating limits. Talen Energy owns the unit, which is H.A. Wagner Generating Station Unit 4. FERC in mid-August approved PJM’s regional cost allocation plan for the Eddystone power plant. In the decision, the agency rejected EKPC’s arguments in that case that the Eddystone costs shouldn’t be allocated across PJM’s footprint. However, earlier this month, PJM returned to FERC, asking the agency to approve regional cost allocation for future DOE section 202(c) orders. The proposal would apply if two conditions are met: the DOE order directs a resource to maintain operations for resource adequacy purposes for the PJM region and isn’t limited to issues in a specific “locational deliverability area” and the power

IT shortcuts curb AI returns

Organizations must ensure the infrastructure is AI ready Infrastructure is another area where Cisco found a major difference. Pacesetters are designing their networks for future demands. Seventy-one percent say their networks can scale instantly for new AI projects. Roughly three-quarters of pacesetters are investing in new data center capacity over the next year. Currently, about two-thirds say their infrastructure can accommodate AI workloads. Most pacesetters (93%) also have data systems that are fully prepared for AI, compared with 34% of other companies. About 76% have fully centralized their in-house data, while only 19% of other companies have done the same. Eighty-four percent report strong governance readiness, while 95% have mature processes to measure the impact of AI. If ever there was a technological shift that requires the right infrastructure, it’s AI. AI generates a significant amount of data, needs large amounts of processes and low latency, high-capacity networks. Historically, businesses could operate with networks that operated on the premise of “best effort,” but that’s no longer the case. From the data center to campus to branch offices, in most companies, the network will require a refresh. Scaling AI requires the right processes When it comes to being disciplined, 62% of pacesetters have an established process for generating, piloting, and scaling AI use cases. Only 13% of other organizations (non-pacesetters) have reached this level of maturity. Most pacesetters say their AI models achieve at least 75% accuracy. Almost half also expect a 50% to 100% return on investment (ROI) within a year, far above the average. Cisco notes that over the past six months, pressure has been building for companies to show tangible ROI. Executives and IT leaders are pushing for results, and so are competitors. By contrast, most other companies are in early stages of readiness. Although 83% plan to

Qualcomm goes all-in on inferencing with purpose-built cards and racks

From a strategy perspective, there is a longer term enterprise play here, noted Moor’s Kimball; Humain is Qualcomm’s first customer, and a cloud service provider (CSP) or hyperscaler will likely be customer number two. However, at some point, these rack-scale systems will find their way into the enterprise. “If I were the AI200 product marketing lead, I would be thinking about how I demonstrate this as a viable platform for those enterprise workloads that will be getting ‘agentified’ over the next several years,” said Kimball. It seems a natural step, as Qualcomm saw success with its AI100 accelerator, a strong inference chip, he noted. Right now, Nvidia and AMD dominate the training market, with CUDA and ROCm enjoying a “stickiness” with customers. “If I am a semiconductor giant like Qualcomm that is so good at understanding the performance-power balance, this inference market makes perfect sense to really lean in on,” said Kimball. He also pointed to the company’s plans to re-enter the datacenter CPU space with its Oryon CPU, which is featured in Snapdragon and loosely based on technology it acquired with its $1.4 billion Nuvia acquisition. Ultimately, Qualcomm’s move demonstrates how wide open the inference market is, said Kimball. The company, he noted, has been very good at choosing target markets and has seen success when entering those markets. “That the company would decide to go more ‘in’ on the inference market makes sense,” said Kimball. He added that, from an ROI perspective, inferencing will “dwarf” training in terms of volume and dollars.

AI data center building boom risks fueling future debt bust, bank warns

However, that’s only one part of the problem. Meeting the power demands of AI data centers will require the energy sector to make large investments. Then there’s data center demand for microprocessors, rare earth elements, and other valuable metals such as copper, which could, in a bust, make data centers the most expensively-assembled unwanted assets in history. “Financial stability consequences of an AI-related asset price fall could arise through multiple channels. If forecasted debt-financed AI infrastructure growth materializes, the potential financial stability consequences of such an event are likely to grow,” warned the BoE blog post. “For companies who depend on the continued demand for massive computational capacity to train and run inference on AI models, an algorithmic breakthrough or other event which challenges that paradigm could cause a significant re-evaluation of asset prices,” it continued. According to Matt Hasan, CEO of AI consultancy aiRESULTS, the underlying problem is the speed with which AI has emerged. “What we’re witnessing isn’t just an incremental expansion, it’s a rush to construct power-hungry, mega-scale data centers,” he told Network World. The dot.com reversal might be the wrong comparison; it dented the NASDAQ and hurt tech investment, but the damage to organizations investing in e-commerce was relatively limited. AI, by contrast, might have wider effects for large enterprises because so many have pinned their business prospects on its potential. “Your reliance on these large providers means you are indirectly exposed to the stability of their debt. If a correction occurs, the fallout can impact the services you rely on,” said Hasan.

Intel sees supply shortage, will prioritize data center technology

“Capacity constraints, especially on Intel 10 and Intel 7 [Intel’s semiconductor manufacturing process], limited our ability to fully meet demand in Q3 for both data center and client products,” said Zinsner, adding that Intel isn’t about to add capacity to Intel 10 and 7 when it has moved beyond those nodes. “Given the current tight capacity environment, which we expect to persist into 2026, we are working closely with customers to maximize our available output, including adjusting pricing and mix to shift demand towards products where we have supply and they have demand,” said Zinsner. For that reason, Zinzner projects that the fourth quarter will be roughly flat versus the third quarter in terms of revenue. “We expect Intel products up modestly sequentially but below customer demand as we continue to navigate supply environment,” said Zinsner. “We expect CCG to be down modestly and PC AI to be up strongly sequentially as we prioritize wafer capacity for server shipments over entry-level client parts.”

How to set up an AI data center in 90 days

“Personally, I think that a brownfield is very creative way to deal with what I think is the biggest problem that we’ve got right now, which is time and speed to market,” he said. “On a brownfield, I can go into a building that’s already got power coming into the building. Sometimes they’ve already got chiller plants, like what we’ve got with the building I’m in right now.” Patmos certainly made the most of the liquid facilities in the old printing press building. The facility is built to handle anywhere from 50 to over 140 kilowatts per cabinet, a leap far beyond the 1–2 kW densities typical of legacy data centers. The chips used in the servers are Nvidia’s Grace Blackwell processors, which run extraordinarily hot. To manage this heat load, Patmos employs a multi-loop liquid cooling system. The design separates water sources into distinct, closed loops, each serving a specific function and ensuring that municipal water never directly contacts sensitive IT equipment. “We have five different, completely separated water loops in this building,” said Morgan. “The cooling tower uses city water for evaporation, but that water never mixes with the closed loops serving the data hall. Everything is designed to maximize efficiency and protect the hardware.” The building taps into Kansas City’s district chilled water supply, which is sourced from a nearby utility plant. This provides the primary cooling resource for the facility. Inside the data center, a dedicated loop circulates a specialized glycol-based fluid, filtered to extremely low micron levels and formulated to be electronically safe. Heat exchangers transfer heat from the data hall fluid to the district chilled water, keeping the two fluids separate and preventing corrosion or contamination. Liquid-to-chip and rear-door heat exchangers are used for immediate heat removal.

INNIO and VoltaGrid: Landmark 2.3 GW Modular Power Deal Signals New Phase for AI Data Centers

Why This Project Marks a Landmark Shift The deployment of 2.3 GW of modular generation represents utility-scale capacity, but what makes it distinct is the delivery model. Instead of a centralized plant, the project uses modular gas-reciprocating “power packs” that can be phased in step with data-hall readiness. This approach allows staged energization and limits the bottlenecks that often stall AI campuses as they outgrow grid timelines or wait in interconnection queues. AI training loads fluctuate sharply, placing exceptional stress on grid stability and voltage quality. The INNIO/VoltaGrid platform was engineered specifically for these GPU-driven dynamics, emphasizing high transient performance (rapid load acceptance) and grid-grade power quality, all without dependence on batteries. Each power pack is also designed for maximum permitting efficiency and sustainability. Compared with diesel generation, modern gas-reciprocating systems materially reduce both criteria pollutants and CO₂ emissions. VoltaGrid markets the configuration as near-zero criteria air emissions and hydrogen-ready, extending allowable runtimes under air permits and making “prime-as-a-service” viable even in constrained or non-attainment markets. 2025: Momentum for Modular Prime Power INNIO has spent 2025 positioning its Jenbacher platform as a next-generation power solution for data centers: combining fast start, high transient performance, and lower emissions compared with diesel. While the 3 MW J620 fast-start lineage dates back to 2019, this year the company sharpened its data center narrative and booked grid stability and peaking projects in markets where rapid data center growth is stressing local grids. This momentum was exemplified by an 80 MW deployment in Indonesia announced earlier in October. The same year saw surging AI-driven demand and INNIO’s growing push into North American data-center markets. Specifications for the 2.3 GW VoltaGrid package highlight the platform’s heat tolerance, efficiency, and transient response, all key attributes for powering modern AI campuses. VoltaGrid’s 2025 Milestones VoltaGrid’s announcements across 2025 reflect

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