Get ONMINE

Stay Ahead, Stay ONMINE

Introducing Gemma 3n: The developer guide

The first Gemma model launched early last year and has since grown into a thriving Gemmaverse of over 160 million collective downloads. This ecosystem includes our family of over a dozen specialized models for everything from safeguarding to medical applications and, most inspiringly, the countless innovations from the community. From innovators like Roboflow building enterprise computer vision to the Institute of Science Tokyo creating highly-capable Japanese Gemma variants, your work has shown us the path forward.Building on this incredible momentum, we’re excited to announce the full release of Gemma 3n. While last month’s preview offered a glimpse, today unlocks the full power of this mobile-first architecture. Gemma 3n is designed for the developer community that helped shape Gemma. It’s supported by your favorite tools including Hugging Face Transformers, llama.cpp, Google AI Edge, Ollama, MLX, and many others, enabling you to fine-tune and deploy for your specific on-device applications with ease. This post is the developer deep dive: we’ll explore some of the innovations behind Gemma 3n, share new benchmark results, and show you how to start building today.What’s new in Gemma 3n?Gemma 3n represents a major advancement for on-device AI, bringing powerful multimodal capabilities to edge devices with performance previously only seen in last year’s cloud-based frontier models. Achieving this leap in on-device performance required rethinking the model from the ground up. The foundation is Gemma 3n’s unique mobile-first architecture, and it all starts with MatFormer.MatFormer: One model, many sizesAt the core of Gemma 3n is the MatFormer (🪆Matryoshka Transformer) architecture, a novel nested transformer built for elastic inference. Think of it like Matryoshka dolls: a larger model contains smaller, fully functional versions of itself. This approach extends the concept of Matryoshka Representation Learning from just embeddings to all transformer components. During the MatFormer training of the 4B effective parameter (E4B) model, a 2B effective parameter (E2B) sub-model is simultaneously optimized within it, as shown in the figure above. This provides developers two powerful capabilities and use cases today:1: Pre-extracted models: You can directly download and use either the main E4B model for the highest capabilities, or the standalone E2B sub-model which we have already extracted for you, offering up to 2x faster inference.2: Custom sizes with Mix-n-Match: For more granular control tailored to specific hardware constraints, you can create a spectrum of custom-sized models between E2B and E4B using a method we call Mix-n-Match. This technique allows you to precisely slice the E4B model’s parameters, primarily by adjusting the feed forward network hidden dimension per layer (from 8192 to 16384) and selectively skipping some layers. We are releasing the MatFormer Lab, a tool that shows how to retrieve these optimal models, which were identified by evaluating various settings on benchmarks like MMLU. MMLU scores for the pre-trained Gemma 3n checkpoints at different model sizes (using Mix-n-Match) Looking ahead, the MatFormer architecture also paves the way for elastic execution. While not part of today’s launched implementations, this capability allows a single deployed E4B model to dynamically switch between E4B and E2B inference paths on the fly, enabling real-time optimization of performance and memory usage based on the current task and device load.Per-Layer Embeddings (PLE): Unlocking more memory efficiencyGemma 3n models incorporate Per-Layer Embeddings (PLE). This innovation is tailored for on-device deployment as it dramatically improves model quality without increasing the high-speed memory footprint required on your device’s accelerator (GPU/TPU).While the Gemma 3n E2B and E4B models have a total parameter count of 5B and 8B respectively, PLE allows a significant portion of these parameters (the embeddings associated with each layer) to be loaded and computed efficiently on the CPU. This means only the core transformer weights (approximately 2B for E2B and 4B for E4B) need to sit in the typically more constrained accelerator memory (VRAM). With Per-Layer Embeddings, you can use Gemma 3n E2B while only having ~2B parameters loaded in your accelerator. KV Cache sharing: Faster long-context processingProcessing long inputs, such as the sequences derived from audio and video streams, is essential for many advanced on-device multimodal applications. Gemma 3n introduces KV Cache Sharing, a feature designed to significantly accelerate time-to-first-token for streaming response applications.KV Cache Sharing optimizes how the model handles the initial input processing stage (often called the “prefill” phase). The keys and values of the middle layer from local and global attention are directly shared with all the top layers, delivering a notable 2x improvement on prefill performance compared to Gemma 3 4B. This means the model can ingest and understand lengthy prompt sequences much faster than before.Audio understanding: Introducing speech to text and translationGemma 3n uses an advanced audio encoder based on the Universal Speech Model (USM). The encoder generates a token for every 160ms of audio (about 6 tokens per second), which are then integrated as input to the language model, providing a granular representation of the sound context.This integrated audio capability unlocks key features for on-device development, including:Automatic Speech Recognition (ASR): Enable high-quality speech-to-text transcription directly on the device.Automatic Speech Translation (AST): Translate spoken language into text in another language.We’ve observed particularly strong AST results for translation between English and Spanish, French, Italian, and Portuguese, offering great potential for developers targeting applications in these languages. For tasks like speech translation, leveraging Chain-of-Thought prompting can significantly enhance results. Here’s an example: user Transcribe the following speech segment in Spanish, then translate it into English: model Plain text At launch time, the Gemma 3n encoder is implemented to process audio clips up to 30 seconds. However, this is not a fundamental limitation. The underlying audio encoder is a streaming encoder, capable of processing arbitrarily long audios with additional long form audio training. Follow-up implementations will unlock low-latency, long streaming applications.MobileNet-V5: New state-of-the-art vision encoderAlongside its integrated audio capabilities, Gemma 3n features a new, highly efficient vision encoder, MobileNet-V5-300M, delivering state-of-the-art performance for multimodal tasks on edge devices.Designed for flexibility and power on constrained hardware, MobileNet-V5 gives developers:Multiple input resolutions: Natively supports resolutions of 256×256, 512×512, and 768×768 pixels, allowing you to balance performance and detail for your specific applications.Broad visual understanding: Co-trained on extensive multimodal datasets, it excels at a wide range of image and video comprehension tasks.High throughput: Processes up to 60 frames per second on a Google Pixel, enabling real-time, on-device video analysis and interactive experiences.This level of performance is achieved with multiple architectural innovations, including:An advanced foundation of MobileNet-V4 blocks (including Universal Inverted Bottlenecks and Mobile MQA).A significantly scaled up architecture, featuring a hybrid, deep pyramid model that is 10x larger than the biggest MobileNet-V4 variant.A novel Multi-Scale Fusion VLM adapter that enhances the quality of tokens for better accuracy and efficiency.Benefiting from novel architectural designs and advanced distillation techniques, MobileNet-V5-300M substantially outperforms the baseline SoViT in Gemma 3 (trained with SigLip, no distillation). On a Google Pixel Edge TPU, it delivers a 13x speedup with quantization (6.5x without), requires 46% fewer parameters, and has a 4x smaller memory footprint, all while providing significantly higher accuracy on vision-language tasksWe’re excited to share more about the work behind this model. Look out for our upcoming MobileNet-V5 technical report, which will deep dive into the model architecture, data scaling strategies, and advanced distillation techniques.Making Gemma 3n accessible from day one has been a priority. We’re proud to partner with many incredible open source developers to ensure broad support across popular tools and platforms, including contributions from teams behind AMD, Axolotl, Docker, Hugging Face, llama.cpp, LMStudio, MLX, NVIDIA, Ollama, RedHat, SGLang, Unsloth, and vLLM.But this ecosystem is just the beginning. The true power of this technology is in what you will build with it. That’s why we’re launching the Gemma 3n Impact Challenge. Your mission: use Gemma 3n’s unique on-device, offline, and multimodal capabilities to build a product for a better world. With $150,000 in prizes, we’re looking for a compelling video story and a “wow” factor demo that shows real-world impact. Join the challenge and help build a better future.Get started with Gemma 3n todayReady to explore the potential of Gemma 3n today? Here’s how:Experiment directly: Use Google AI Studio to try Gemma 3n in just a couple of clicks. Gemma models can also be deployed directly to Cloud Run from AI Studio.Learn & integrate: Dive into our comprehensive documentation to quickly integrate Gemma into your projects or start with our inference and fine-tuning guides.

The first Gemma model launched early last year and has since grown into a thriving Gemmaverse of over 160 million collective downloads. This ecosystem includes our family of over a dozen specialized models for everything from safeguarding to medical applications and, most inspiringly, the countless innovations from the community. From innovators like Roboflow building enterprise computer vision to the Institute of Science Tokyo creating highly-capable Japanese Gemma variants, your work has shown us the path forward.

Building on this incredible momentum, we’re excited to announce the full release of Gemma 3n. While last month’s preview offered a glimpse, today unlocks the full power of this mobile-first architecture. Gemma 3n is designed for the developer community that helped shape Gemma. It’s supported by your favorite tools including Hugging Face Transformers, llama.cpp, Google AI Edge, Ollama, MLX, and many others, enabling you to fine-tune and deploy for your specific on-device applications with ease. This post is the developer deep dive: we’ll explore some of the innovations behind Gemma 3n, share new benchmark results, and show you how to start building today.


What’s new in Gemma 3n?

Gemma 3n represents a major advancement for on-device AI, bringing powerful multimodal capabilities to edge devices with performance previously only seen in last year’s cloud-based frontier models.

Achieving this leap in on-device performance required rethinking the model from the ground up. The foundation is Gemma 3n’s unique mobile-first architecture, and it all starts with MatFormer.

MatFormer: One model, many sizes

At the core of Gemma 3n is the MatFormer (🪆Matryoshka Transformer) architecture, a novel nested transformer built for elastic inference. Think of it like Matryoshka dolls: a larger model contains smaller, fully functional versions of itself. This approach extends the concept of Matryoshka Representation Learning from just embeddings to all transformer components.

During the MatFormer training of the 4B effective parameter (E4B) model, a 2B effective parameter (E2B) sub-model is simultaneously optimized within it, as shown in the figure above. This provides developers two powerful capabilities and use cases today:

1: Pre-extracted models: You can directly download and use either the main E4B model for the highest capabilities, or the standalone E2B sub-model which we have already extracted for you, offering up to 2x faster inference.

2: Custom sizes with Mix-n-Match: For more granular control tailored to specific hardware constraints, you can create a spectrum of custom-sized models between E2B and E4B using a method we call Mix-n-Match. This technique allows you to precisely slice the E4B model’s parameters, primarily by adjusting the feed forward network hidden dimension per layer (from 8192 to 16384) and selectively skipping some layers. We are releasing the MatFormer Lab, a tool that shows how to retrieve these optimal models, which were identified by evaluating various settings on benchmarks like MMLU.

Custom Sizes with Mix-n-Match

MMLU scores for the pre-trained Gemma 3n checkpoints at different model sizes (using Mix-n-Match)

Looking ahead, the MatFormer architecture also paves the way for elastic execution. While not part of today’s launched implementations, this capability allows a single deployed E4B model to dynamically switch between E4B and E2B inference paths on the fly, enabling real-time optimization of performance and memory usage based on the current task and device load.

Per-Layer Embeddings (PLE): Unlocking more memory efficiency

Gemma 3n models incorporate Per-Layer Embeddings (PLE). This innovation is tailored for on-device deployment as it dramatically improves model quality without increasing the high-speed memory footprint required on your device’s accelerator (GPU/TPU).

While the Gemma 3n E2B and E4B models have a total parameter count of 5B and 8B respectively, PLE allows a significant portion of these parameters (the embeddings associated with each layer) to be loaded and computed efficiently on the CPU. This means only the core transformer weights (approximately 2B for E2B and 4B for E4B) need to sit in the typically more constrained accelerator memory (VRAM).

Per-Layer Embeddings

With Per-Layer Embeddings, you can use Gemma 3n E2B while only having ~2B parameters loaded in your accelerator.

KV Cache sharing: Faster long-context processing

Processing long inputs, such as the sequences derived from audio and video streams, is essential for many advanced on-device multimodal applications. Gemma 3n introduces KV Cache Sharing, a feature designed to significantly accelerate time-to-first-token for streaming response applications.

KV Cache Sharing optimizes how the model handles the initial input processing stage (often called the “prefill” phase). The keys and values of the middle layer from local and global attention are directly shared with all the top layers, delivering a notable 2x improvement on prefill performance compared to Gemma 3 4B. This means the model can ingest and understand lengthy prompt sequences much faster than before.

Audio understanding: Introducing speech to text and translation

Gemma 3n uses an advanced audio encoder based on the Universal Speech Model (USM). The encoder generates a token for every 160ms of audio (about 6 tokens per second), which are then integrated as input to the language model, providing a granular representation of the sound context.

This integrated audio capability unlocks key features for on-device development, including:

  • Automatic Speech Recognition (ASR): Enable high-quality speech-to-text transcription directly on the device.
  • Automatic Speech Translation (AST): Translate spoken language into text in another language.

We’ve observed particularly strong AST results for translation between English and Spanish, French, Italian, and Portuguese, offering great potential for developers targeting applications in these languages. For tasks like speech translation, leveraging Chain-of-Thought prompting can significantly enhance results. Here’s an example:

user
Transcribe the following speech segment in Spanish, then translate it into English: 

model

Plain text

At launch time, the Gemma 3n encoder is implemented to process audio clips up to 30 seconds. However, this is not a fundamental limitation. The underlying audio encoder is a streaming encoder, capable of processing arbitrarily long audios with additional long form audio training. Follow-up implementations will unlock low-latency, long streaming applications.


MobileNet-V5: New state-of-the-art vision encoder

Alongside its integrated audio capabilities, Gemma 3n features a new, highly efficient vision encoder, MobileNet-V5-300M, delivering state-of-the-art performance for multimodal tasks on edge devices.

Designed for flexibility and power on constrained hardware, MobileNet-V5 gives developers:

  • Multiple input resolutions: Natively supports resolutions of 256×256, 512×512, and 768×768 pixels, allowing you to balance performance and detail for your specific applications.
  • Broad visual understanding: Co-trained on extensive multimodal datasets, it excels at a wide range of image and video comprehension tasks.
  • High throughput: Processes up to 60 frames per second on a Google Pixel, enabling real-time, on-device video analysis and interactive experiences.

This level of performance is achieved with multiple architectural innovations, including:

  • An advanced foundation of MobileNet-V4 blocks (including Universal Inverted Bottlenecks and Mobile MQA).
  • A significantly scaled up architecture, featuring a hybrid, deep pyramid model that is 10x larger than the biggest MobileNet-V4 variant.
  • A novel Multi-Scale Fusion VLM adapter that enhances the quality of tokens for better accuracy and efficiency.

Benefiting from novel architectural designs and advanced distillation techniques, MobileNet-V5-300M substantially outperforms the baseline SoViT in Gemma 3 (trained with SigLip, no distillation). On a Google Pixel Edge TPU, it delivers a 13x speedup with quantization (6.5x without), requires 46% fewer parameters, and has a 4x smaller memory footprint, all while providing significantly higher accuracy on vision-language tasks

We’re excited to share more about the work behind this model. Look out for our upcoming MobileNet-V5 technical report, which will deep dive into the model architecture, data scaling strategies, and advanced distillation techniques.

Making Gemma 3n accessible from day one has been a priority. We’re proud to partner with many incredible open source developers to ensure broad support across popular tools and platforms, including contributions from teams behind AMD, Axolotl, Docker, Hugging Face, llama.cpp, LMStudio, MLX, NVIDIA, Ollama, RedHat, SGLang, Unsloth, and vLLM.

But this ecosystem is just the beginning. The true power of this technology is in what you will build with it. That’s why we’re launching the Gemma 3n Impact Challenge. Your mission: use Gemma 3n’s unique on-device, offline, and multimodal capabilities to build a product for a better world. With $150,000 in prizes, we’re looking for a compelling video story and a “wow” factor demo that shows real-world impact. Join the challenge and help build a better future.

Get started with Gemma 3n today

Ready to explore the potential of Gemma 3n today? Here’s how:

  • Experiment directly: Use Google AI Studio to try Gemma 3n in just a couple of clicks. Gemma models can also be deployed directly to Cloud Run from AI Studio.
  • Learn & integrate: Dive into our comprehensive documentation to quickly integrate Gemma into your projects or start with our inference and fine-tuning guides.
Shape
Shape
Stay Ahead

Explore More Insights

More Articles

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.

Shape

AI shifts IT roles from operator to orchestrator

The report indicates that IT roles are becoming more strategic and automation-driven, with 52% of respondents citing increases in both areas. Roles are also becoming more cross-functional (47%) and complex (41%), reflecting the integration of AI into broader business processes. AI is also affecting how IT teams allocate time. Respondents

Read More »

Energy Department Awards New Contracts from Strategic Petroleum Reserve, Advancing Emergency Exchange

WASHINGTON—The U.S. Department of Energy’s (DOE) Hydrocarbons and Geothermal Energy Office (HGEO) today announced awards of contracts to exchange 26 million barrels of crude oil from the Strategic Petroleum Reserve (SPR) at the West Hackberry site, marking the next phase of DOE’s execution of the United States’ 172-million-barrel contribution to the International Energy Agency’s collective action to stabilize global oil supply. These awards follow DOE’s recent Request for Proposal (RFP) for this portion of the emergency exchange, with deliveries beginning immediately as the Department continues to move quickly to address short-term supply disruptions and strengthen energy security for the United States. “Through this emergency exchange, the Department is taking swift action to support near‑term supply needs while strengthening the Strategic Petroleum Reserve for the long term,” said Kyle Haustveit, Assistant Secretary of the Hydrocarbons and Geothermal Energy Office. “By returning additional premium barrels at no cost to taxpayers, this exchange reinforces market reliability today and delivers meaningful value to the American people when those barrels are returned.” Under these awards, DOE will move forward with an exchange of 26 million barrels of crude oil, which will be returned with additional premium barrels by next year—supporting energy security and delivering value for the American people at no cost to taxpayers. This action builds on earlier exchange actions, which have already awarded approximately 55 million barrels from the Bayou Choctaw, Bryan Mound, and West Hackberry sites, demonstrating the reserve’s ability to deliver crude efficiently under emergency conditions. To date, more than 10 million barrels have already been delivered to market. The exchange also allows participating companies to take advantage of the President’s limited Jones Act waiver, helping accelerate critical near-term oil flows into the market. Companies can begin scheduling deliveries immediately. DOE will continue to evaluate market conditions and operational capacity as it advances

Read More »

Apply Now: 2026 Waste to Energy and Materials Technical Assistance for State, Local, and Tribal Governments

The U.S. Department of Energy’s Alternative Fuels and Feedstocks Office (AFFO), formerly known as the Bioenergy Technologies Office, and the National Laboratory of the Rockies (NLR) are launching the 2026 Waste to Energy and Materials Technical Assistance Program for state, local, and Tribal governments. The scope of this year’s program has been expanded to include additional municipal solid waste materials such as electronics, industrial wastewater, and other byproducts.  U.S. waste streams present significant logistical and economic challenges for states, counties, municipalities, and Tribal governments. However, waste is also a resource that can be used as an unconventional additional source of energy, advanced materials, and critical minerals. This program provides no-cost technical assistance to states, counties, municipalities, and Tribal governments with the most relevant data to guide decision-making—providing local solutions to the various aspects of waste management, taking into consideration current handling practices, costs, and infrastructure. It is designed to help officials evaluate the most sensible end uses for their waste, whether repurposing it for on-site heat and power, upgrading it into transportation fuels, or using it for material and mineral recovery. Program technical assistance includes: Waste resource information Infrastructure considerations Techno-economic comparison of energy, material, and mineral recovery options Evaluation and sharing of case studies (to the extent possible) from similar communities/projects The 2026 Waste to Energy and Materials Technical Assistance application portal is now open and applications will be accepted through May 30, 2026. For information on applicant eligibility and how to apply, please visit NLR’s technical assistance webpage. Timeline for Technical Assistance Opportunity Date Action April 15, 2026 Application Portal Opens May 30, 2026 Application Portal Closes  July – August 2026 Selections Made and Recipients Informed  Learn more about AFFO-supported waste to energy and materials technical assistance. If you have further questions, please see frequently asked questions or contact the Waste to

Read More »

Energy Deputy Secretary Danly Commends FERC Action on Large Load Interconnection Reform

WASHINGTON—U.S. Deputy Secretary of Energy James P. Danly issued the following statement after the Federal Energy Regulatory Commission (FERC or Commission) announced it will take action by June 2026 on the large load interconnection proceeding initiated at the direction of U.S. Secretary of Energy Chris Wright: “FERC’s announcement today demonstrates Chairman Swett’s commitment to implement Secretary Wright’s directive that the Commission ensure the timely and orderly integration of large electric loads that deliver on President Trump’s goal of American energy dominance. “I expect that the Commission will act quickly and decisively to improve interconnection processes, support the co-location of load and generation, and accelerate the addition of new generation to ensure that supply is built alongside demand—delivering affordable, reliable, and secure energy for all Americans. “Having served at FERC as commissioner and chairman, I understand FERC’s role in ensuring the reliability of the nation’s bulk power system, and I commend Chairman Swett for focusing on affordability and reliability.”                                                                                               ###  

Read More »

Petrobras discovers hydrocarbons in Campos basin presalt offshore Brazil

@import url(‘https://fonts.googleapis.com/css2?family=Inter:[email protected]&display=swap’); .ebm-page__main h1, .ebm-page__main h2, .ebm-page__main h3, .ebm-page__main h4, .ebm-page__main h5, .ebm-page__main h6 { font-family: Inter; } body { line-height: 150%; letter-spacing: 0.025em; } button, .ebm-button-wrapper { font-family: Inter; } .label-style { text-transform: uppercase; color: var(–color-grey); font-weight: 600; font-size: 0.75rem; } .caption-style { font-size: 0.75rem; opacity: .6; } #onetrust-pc-sdk [id*=btn-handler], #onetrust-pc-sdk [class*=btn-handler] { background-color: #c19a06 !important; border-color: #c19a06 !important; } #onetrust-policy a, #onetrust-pc-sdk a, #ot-pc-content a { color: #c19a06 !important; } #onetrust-consent-sdk #onetrust-pc-sdk .ot-active-menu { border-color: #c19a06 !important; } #onetrust-consent-sdk #onetrust-accept-btn-handler, #onetrust-banner-sdk #onetrust-reject-all-handler, #onetrust-consent-sdk #onetrust-pc-btn-handler.cookie-setting-link { background-color: #c19a06 !important; border-color: #c19a06 !important; } #onetrust-consent-sdk .onetrust-pc-btn-handler { color: #c19a06 !important; border-color: #c19a06 !important; } Petrobras has discovered presence in the Campos basin presalt offshore Brazil during exploration in sector SC-AP4, block CM-477. Samples taken from the well, 1-BRSA-1404DC-RJS, will be sent for laboratory analysis with the aim of characterizing the conditions of the reservoirs and fluids found to enable continued evaluation of the area’s potential, the company said in a release Apr. 13. The discovery well was drilled 201 km off the coast of the state of Rio de Janeiro in water depth of 2,984 m. The hydrocarbon-bearing interval was confirmed through electrical profiles, gas evidence, and fluid sampling. Petrobras is the operator of block CM-477 with 70% interest. bp plc holds the remaining 30%.

Read More »

bp to operate blocks offshore Namibia through acquisition

@import url(‘https://fonts.googleapis.com/css2?family=Inter:[email protected]&display=swap’); .ebm-page__main h1, .ebm-page__main h2, .ebm-page__main h3, .ebm-page__main h4, .ebm-page__main h5, .ebm-page__main h6 { font-family: Inter; } body { line-height: 150%; letter-spacing: 0.025em; } button, .ebm-button-wrapper { font-family: Inter; } .label-style { text-transform: uppercase; color: var(–color-grey); font-weight: 600; font-size: 0.75rem; } .caption-style { font-size: 0.75rem; opacity: .6; } #onetrust-pc-sdk [id*=btn-handler], #onetrust-pc-sdk [class*=btn-handler] { background-color: #c19a06 !important; border-color: #c19a06 !important; } #onetrust-policy a, #onetrust-pc-sdk a, #ot-pc-content a { color: #c19a06 !important; } #onetrust-consent-sdk #onetrust-pc-sdk .ot-active-menu { border-color: #c19a06 !important; } #onetrust-consent-sdk #onetrust-accept-btn-handler, #onetrust-banner-sdk #onetrust-reject-all-handler, #onetrust-consent-sdk #onetrust-pc-btn-handler.cookie-setting-link { background-color: #c19a06 !important; border-color: #c19a06 !important; } #onetrust-consent-sdk .onetrust-pc-btn-handler { color: #c19a06 !important; border-color: #c19a06 !important; } Map from bp plc <!–> –> bp plc aims to become operator of three exploration blocks offshore Namibia through acquisition of a 60% interest from Eco Atlantic Oil & Gas. Subject to Namibian government and joint venture partner approvals, bp will operate blocks PEL97, PEL99, and PEL100 in Walvis basin.   In a release Apr. 13, bp said entering the blocks builds on its recent exploration successes in Namibia through Azule Energy, a 50-50 joint venture between bp and Eni. Eco Atlantic will remain a partner, along with Namibia’s national oil company NAMCOR, following the deal’s closing, which is subject to closing conditions.

Read More »

ConocoPhillips sends team to Venezuela to evaluate oil, gas opportunities

ConocoPhillips sent a team to Venezuela to evaluate oil and gas opportunities, the company confirmed to Oil & Gas Journal Apr. 13. In an email to OGJ, a company spokesperson said “ConocoPhillips can confirm that we sent a small evaluation team to Venezuela during the week of Apr. 6 to better understand the potential for in-country oil and gas opportunities.” Asked what clarity the company seeks, the spokesperson said the team “will evaluate Venezuela against other international opportunities as part of our disciplined investment framework.” The operator left Venezuela in 2007 after then-President Hugo Chavez’s government reverted privately run oil fields to state control. ConocoPhillips, along with ExxonMobil, refused the government’s terms and took claims to the World Bank’s International Centre for the Settlement of Investment Disputes (ICSID). ConocoPhillips is owed about $12 billion following two judgements, an amount still sought by the company, which, prior to the expropriation of its interests, held a 50.1% interest in Petrozuata, a 40% interest in Hamaca, and a 32.5% interest in Corocoro heavy oil projects in Venezuela. In January, following the removal of Venezuela’s leader Nicolas Maduro, US President Donald Trump urged oil and gas companies to spend billions to rebuild Venezuela’s energy sector. ExxonMobil, which also exited the country in 2007, ​sent a technical team to Venezuela in March to ⁠evaluate the infrastructure and investment opportunities. In a discussion at CERAWeek by S&P Global in Houston in March, ConocoPhillips’ chief executive officer, Ryan Lance, said Venezuela needs to “completely rewire” ​its fiscal system to attract new ‌investment. The South American country holds a large cache of proven oil reserves, but has faced decades of production challenges due to mismanagement, underinvestment, and sanctions.

Read More »

Data centers are costing local governments billions

Tax benefits for hyperscalers and other data center operators are costing local administrations billions of dollars. In the US, three states are already giving away more than $1 billion in potential tax revenue, while 14 are failing to declare how much data center subsidies are costing taxpayers, according to Good Jobs First. The campaign group said the failure to declare the tax subsidies goes against US Generally Accepted Accounting Principles (GAAP) and that they should, since 2017, be declared as lost revenue. “Tax-abatement laws written long ago for much smaller data centers, predating massive artificial intelligence (AI) facilities, are now unexpectedly costing governments billions of dollars in lost tax revenue,” Good Jobs First said. “Three states, Georgia, Virginia, and Texas, already lose $1 billion or more per year,” it reported in its new study, “Data Center Tax Abatements: Why States and Localities Must Disclose These Soaring Revenue Losses.”

Read More »

Equinix offering targets automated AI-centric network operations

Another component, Fabric Application Connect, functions as a private, dedicated connectivity marketplace for AI services. It lets enterprises access inference, training, storage, and security providers over private connections, bypassing the public Internet and limiting data exposure during AI development and deployment. Operational visibility is provided through Fabric Insights, an AI-powered monitoring layer that analyzes real-time network telemetry to detect anomalies and predict potential issues before they impact workloads. Fabric Insights integrates with security information and event management (SIEM) platforms such as Splunk and Datadog and feeds data directly into Fabric Super-Agent to support automated remediation. Fabric Intelligence operates on top of Equinix’s global infrastructure footprint, which includes hundreds of data centers across dozens of metropolitan markets. The platform is positioned as part of Equinix Fabric, a connectivity portfolio used by thousands of customers worldwide to link cloud providers, enterprises, and network services. Fabric Intelligence is available now to preview.

Read More »

Blue Owl Builds a Capital Platform for the Hyperscale AI Era

Capital as a Service: The Hyperscaler Shift This is not just another project financing. It points to a model in which hyperscalers can externalize a significant portion of the capital required for AI campuses while retaining operational control. Under the Hyperion structure, Meta provides construction and property management, while Blue Owl supplies capital at scale alongside infrastructure expertise. Reuters described the transaction as Meta’s largest private capital deal to date, with the campus projected to exceed 2 gigawatts of capacity. For Blue Owl, it marks a shift in role: from backing developers serving hyperscalers to working directly with a hyperscaler to structure ownership more efficiently at scale. Hyperion also helps explain why this model is gaining traction. Hyperscalers are now deploying capital at a pace that makes flexibility a strategic priority. Structures like the Meta–Blue Owl JV allow them to continue expanding infrastructure without fully absorbing the balance-sheet impact of each new campus. Analyst commentary cited by Reuters suggested the arrangement could help Meta mitigate risk and avoid concentrating too much capital in land, buildings, and long-lived infrastructure, preserving capacity for additional facilities and ongoing AI investment. That is the service Blue Owl is effectively providing. Not just capital, but balance-sheet flexibility at a time when AI infrastructure demand is stretching even the largest technology companies. With major tech firms projected to spend hundreds of billions annually on AI infrastructure, that capability is becoming central to how the next generation of campuses gets built. The Capital Baseline Resets In early 2026, hyperscalers effectively reset the capital baseline for the sector. Alphabet projected $175 billion to $185 billion in annual capex, citing continued constraints across servers, data centers, and networking. Amazon pointed to roughly $200 billion, up from $131 billion the prior year, while noting persistent demand pressure in AWS. Meta

Read More »

OpenAI pulls out of a second Stargate data center deal

“OpenAI is embattled on several fronts. Anthropic has been doing very well in the enterprise, and OpenAI’s cash burn might be a problem if it wants to go public at an astronomical $800 billion+ valuation. This is especially true with higher energy prices due to geopolitics, and the public and regulators increasingly skeptical of AI companies, especially outside of the United States,” Roberts said. “I see these moves as OpenAI tightening its belt a bit and being more deliberate about spending as it moves past the interesting tech demo stage of its existence and is expected to provide a real return for investors.” He added, “I expect it’s a symptom of a broader problem, which is that OpenAI has thrown some good money after bad in bets that didn’t work out, like the Sora platform it just shut down, and it’s under increasing pressure to translate its first-mover advantage into real upside for its investors. Spending operational money instead of capital money might give it some flexibility in the short term, and perhaps that’s what this is about.” All in all, he noted, “on a scale of business-ending event to nothingburger, I would put it somewhere in the middle, maybe a little closer to nothingburger.” Acceligence CIO Yuri Goryunov agreed with Roberts, and said, “OpenAI has a problem with commercialization and runaway operating costs, for sure. They are trying to rightsize their commitments and make sure that they deliver on their core products before they run out of money.” Goryunov described OpenAI’s arrangement with Microsoft in Norway as “prudent financial engineering” that allows it to access the data center resources without having to tie up too much capital. “It’s financial discipline. OpenAI [executives] are starting to behave like grownups.” Forrester senior analyst Alvin Nguyen echoed those thoughts. 

Read More »

DCF Tours: SDC Manhattan, 375 Pearl St.

Power: Redundant utility design in a power-constrained market The tour made equally clear that in Manhattan, power is still the central gating factor. The brochure describes SDC Manhattan as offering 18MW of aggregate power delivered to the building, backed by redundant electrical and mechanical systems, backup generators, and Tier III-type concurrent maintainability. The December 2025 press release updated that picture in a more market-facing way, noting that Sabey is one of the only colocation providers in Manhattan with available power, including nearly a megawatt of turnkey power and 7MW of utility power across two powered shell spaces. Bajrushi’s explanation of the electrical topology helped show how Sabey has made that possible. Standing on the third floor, he described a ring bus tying together four Con Edison feeds. Bajrushi said the feeds all originate from the same substation but take different paths into the building, creating redundancy outside the building as well as within it. He added that if one feed fails, the ring bus remains unaffected, and that only one feed is needed to power everything currently in operation. He also noted that Sabey has the ability to add two more feeds in the future if expansion calls for it. That matters in a city where available utility capacity is hard to come by and where many data center conversations end not with square footage but with a megawatt number. Bajrushi also noted that physical space is not the core constraint at 375 Pearl. He said the building still has plenty of room for future buildouts, including open areas that could become additional white space, chiller capacity, or other infrastructure. The bigger question, he suggested, is how and when power and supporting systems get installed. That observation aligns neatly with Sabey’s press release. The company is effectively arguing that SDC

Read More »

Maine to put brakes on big data centers as AI expansion collides with power limits

Mills has pushed for an exemption protecting a proposed $550 million project at the former Androscoggin paper mill in Jay, arguing it would reuse existing infrastructure without straining the grid. Lawmakers rejected that exemption. Mills’ office did not immediately respond to a request for comment. A national wave, an unanswered federal question Maine is one of at least 12 states now weighing moratorium or restraint legislation, alongside more than 300 data center bills filed across 30-plus states in the current session, according to legislative tracking firm MultiState. The shared concern is energy cost. Data centers could consume up to 12% of total US electricity by 2028, according to the US Department of Energy. On March 25, Senator Bernie Sanders and Alexandria Ocasio-Cortez introduced the AI Data Center Moratorium Act in Congress, which would impose a nationwide freeze on all new data center construction until Congress passes AI safety legislation. The Trump administration has pursued a different path from the legislative approach being taken in states. On March 4, Amazon, Google, Meta, Microsoft, OpenAI, Oracle, and xAI signed the White House’s Ratepayer Protection Pledge, a voluntary commitment by hyperscalers to fund their own power generation rather than pass grid costs to ratepayers. The pledge, published in the Federal Register on March 9, carries no penalties for noncompliance or auditing requirements.

Read More »

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.

Read More »

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

Read More »

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

Read More »

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

Read More »
More Articles