AI Infrastructure and Compute
The hardware and cloud stack behind AI, including GPUs, accelerators, data centers, networking, inference chips, and compute supply constraints.
Meeting Rising Power Demand Requires More Than Building Generation
At Aspen Ideas: Economy in Tulsa, ONE Gas CEO Sid McAnnally argued that meeting rising electricity demand requires treating energy sources as complements, not rivals—and distinguishing generation from the infrastructure needed to deliver power affordably and reliably. Commonwealth Fusion Systems CFO Lorence Kim and Cornerstone Government Affairs principal Mark Yates added that the choices involve more than technology: grid connections, local economic benefits and policy stability will shape which projects get built and who benefits.
GPU Forward Curves Point to Rising H100 and B200 Prices
Carmen Li, CEO of Compute Exchange and Silicon Data, argues that buyers should procure GPU capacity through a neutral exchange rather than negotiate separately with hundreds of providers. Drawing on her experience as an energy-options market maker, she says an exchange can verify buyers and sellers, compare offers and support reserved and forward contracts without owning the GPUs it trades. Li also points to Silicon Data’s forward curves, which she says indicate that the market expects H100 and B200 prices to rise.
Europe’s AI Strategy Depends on Building Local Compute and AI Factories
In his GTC Berlin keynote, NVIDIA CEO Jensen Huang argued that Europe’s AI economy will depend on producing intelligence, not just adopting AI services. He called for investment across the stack—from energy and computing to AI factories, models and applications—and presented NVIDIA’s hardware, software and partner ecosystem as infrastructure for that production. His case extended from industrial agents and robotics to secure runtimes and local workstations: AI, he said, requires systems built to run models, connect them to knowledge and tools, and put them to work.
Meta’s Planned Capital Raise Looms Over AI Spending and Shares
Meta’s new nonvoting share class could give Mark Zuckerberg liquidity to cover a tax bill if California’s proposed wealth tax passes, while preserving his control, Ed Ludlow said. Alison Porter, a portfolio manager at Janus Henderson, said investors had anticipated a capital raise to support Meta’s growth spending and that uncertainty over its timing will weigh on the stock; without a third-party cloud business, she added, investors have less visibility into what the company’s AI spending will yield.
GPU Contracts Assume Steady Demand; AI Workloads Do Not
Modal CEO Erik Bernhardsson argues that AI workloads need compute on demand, while much of the GPU market sells capacity in fixed, multiyear blocks that may begin months later. Modal pools supply across customers, regions and providers to match that capacity to demand that rises and falls. Bernhardsson says the model can make flexible access work, but leaves Modal exposed to the financial risk of long-term commitments if GPU prices or demand fall.
AI Demand Fluctuates, but GPU Capacity Still Sells in Three-Year Blocks
Modal CEO Erik Bernhardsson argues that AI workloads fluctuate too much for the GPU market’s prevailing model of capacity sold in fixed, multi-year blocks. In his account, Modal pools supply across customers and scales it with demand, but that flexibility requires the company to make increasingly large commitments to cloud providers—leaving it exposed if GPU prices fall or utilization weakens. The tension is central to how Bernhardsson weighs growth against the financial risks of securing capacity.
Open-Weight Models Could Capture Most Enterprise AI Workloads
Arvind Krishna and Nikesh Arora argue that companies should prepare for major technology shifts without treating every new capability as ready for every job. Krishna says organizations need to begin migrating to post-quantum cryptography before quantum computers can break current encryption; Arora says AI models should be chosen according to the task’s cost and the consequences of failure. Their discussion of cyber defense and software replacement makes the same qualification: adoption can bring advantages, but reliability sets limits.
A SpaceX–Tesla Tie-Up Would Unite Production From Chips to AI
Chad Anderson, an early SpaceX investor and CEO of Space Capital, says a potential merger with Tesla would bring together companies already converging operationally, creating a vertically integrated business that could span chip fabrication through AI services. He compares SpaceX’s role in building infrastructure for intelligence with the industrial giants of the early 20th century, but says the growth case depends on Starship becoming reusable: more launch capacity is needed to expand connectivity and AI computing in orbit.
NVIDIA’s GTC Berlin Keynote Frames Tokens as a Path From Data to Action
NVIDIA CEO Jensen Huang’s GTC Berlin keynote opening video presents tokens as the building blocks that turn data into knowledge and action. Its examples range from treatment planning and crowd forecasts to robots carrying medical samples and inspecting hazardous sites, connecting AI to decisions and physical work. The video places those examples alongside computing infrastructure and closes with a shared ambition for a future built for all humankind.
Micro Duck’s Walking Policy Trains in the Cloud, Then Returns to Local Simulation
The workflow uses NVIDIA RTX Spark for scene creation and policy evaluation, while cloud compute handles parallel training in Isaac Lab; the trained policy then returns to the local system before deployment to Micro Duck. The operator says Astra uses tools including FreeCAD, Blender and Isaac Sim to build the course and robot model, and that Isaac Lab trains thousands of Micro Ducks in parallel. The account shows the robot walking after deployment but reports no separate measurement of its real-world performance.
Factory-Tested AI Pods Could Cut On-Site Deployment From Months to Days
NVIDIA presents DSX as a full-stack reference architecture for designing AI factories as coordinated systems, with compute, networking, power and cooling validated in a digital twin before construction. The company says its DSX Max LPS can deliver the same output with 30% less power, while closed-loop cooling and outdoor dry coolers can save up to 54 million liters of water a day at a 2-gigawatt scale in suitable climates. Its DSX POD packages pre-integrated, factory-tested equipment to cut on-site deployment from months to days once utilities are ready.
NVIDIA Frames AI Compute as Revenue-Producing Infrastructure
NVIDIA’s GTC Berlin 2026 keynote closing video presents AI computing as industrial infrastructure with the potential to generate revenue, linking GPUs and data centers to a wider stack of models, software and robotics. In a musical recap performed by robot characters, the company also calls for agents to act within security limits and urges Europe to build across that stack. The performance makes these points through slogans rather than technical or financial evidence.
Google’s Frontier AI Push Is Constrained by Compute Demand
Google CEO Sundar Pichai argues that staying at the frontier of AI is a strategic necessity, even as the models are costly, compute is scarce and many products can use smaller systems. In an interview with Big Technology Podcast, he said Google is prioritizing research and model development over some short-term Cloud revenue, while planning capacity years ahead. He also described a broader strategy: develop models and agents across Google’s products, and invest in safeguards as AI capabilities advance.
Nscale’s $45 Billion Anthropic Deal Hinges on Building Its West Virginia Campus
Anthropic is in talks with Nscale on a computing deal that could reach $45 billion, for a second facility at Nscale’s Monarch campus in West Virginia. Bloomberg News reporter Bailey Lipschultz said the proposed facility would supply 460 megawatts of compute, but the deal also sharpens the central question facing Nscale: whether it can build the capacity on schedule and deliver the compute underpinning its growth and valuation plans.
SpaceX’s $100 Billion Revenue Goal Puts Capital Spending in Focus
Ahead of SpaceX’s third-quarter results, Bloomberg Intelligence’s Mandeep Singh said investors would focus on whether the company can finance the infrastructure needed to meet its growth targets. He cited the prospect of capital spending above $150 billion in 2027, arguing that contracted demand and guidance for $100 billion in annual recurring revenue do not settle whether investors will accept the cost of building capacity. George Ferguson said launches mainly enable businesses such as Starlink and future space-based data centers, while Singh described revenue from Grok itself as less certain than income from renting compute.
AI Data-Center Forecast Rises to 50 Gigawatts as Demand Economics Remain Unsettled
John Coogan and Jordi Hays weigh AI’s infrastructure forecasts against the harder question of whether businesses can turn the capacity into economic value: SemiAnalysis raised its 2027 data-center forecast to 50 gigawatts, while the hosts said demand and spending remain uncertain. They apply a similar test to claims about Theranos, arguing that Holmes’s story should be judged against evidence of execution in an established medical-device industry. In their account of Poppi’s sale to PepsiCo, the headline $2.2 billion included $300 million contingent on milestones; Pepsi announced the earnout would be $12 million.
Anthropic’s Revenue Growth Slowed as IPO Expectations Rose
Anthropic reported $16.8 billion in third-quarter revenue, up 45% from the previous quarter—a sharp slowdown from the more than 140% growth recorded from the first quarter to the second during the Claude Code surge. Ranjan Roy of Margins and journalist Alex Kantrowitz argue that the figures are hard to interpret without full financial disclosures, especially as investors weigh the company’s growth against expectations for its valuation. Roy says releasing Anthropic’s S-1 could help distinguish company figures from estimates and partial reports.
A Predictable Permitting Process Could Restore Confidence in U.S. Construction
The question Lisa Epifani posed is whether the United States’ institutions can still support prosperity and technological change by enabling the country to build. Oklahoma Sen. Alan Armstrong argued that permitting reform should make approvals clear and predictable, while ensuring projects can proceed once permits are lawfully issued. He said that certainty would make it possible to build while continuing to meet environmental requirements.
Roadster Unveiling Puts Tesla–SpaceX Merger Speculation on Display
Tesla’s Roadster unveiling at a SpaceX facility made visible a possible connection between the companies as investors weigh whether they could merge, Bloomberg Intelligence analyst Matthew Bloxham said. He called the display part of the story Elon Musk has been building around a tie-up, pointing to existing collaboration and potential overlap in data and AI. But Bloxham noted that the Roadster is not expected to be available until 2028, and Tesla has made no official announcement about a merger.
Data-Center Builders Borrowing at 9% Risk Destroying Value
DA Davidson technology research chief Gil Luria argues that AI demand and the risks of financing its infrastructure should be judged separately. He sees chip demand and Nvidia’s ecosystem supporting growth through 2030, but warns that data-center builders borrowing at high rates may destroy value even as demand rises.
The $31 Trillion AI Buildout Depends on Power, Politics and Public Consent
PwC’s Dallas Dolen argues that global AI infrastructure investment could reach $31 trillion through 2050 because demand for compute is likely to outpace efficiency gains. He says the forecast depends not only on data centers and chips but on new power supply, financing and government approvals—and on whether companies can win community support for projects that raise questions about local costs and resources. Dolen expects political and economic disruptions to shift the timing of construction more than erase the underlying demand.
RTX 5090 Outruns M5 Ultra on Models That Fit in VRAM
A comparison of Apple’s M5 Ultra Mac Studio and NVIDIA’s RTX 5090 finds the graphics card faster on a Qwen3.8-27B test, leading in prompt processing and token generation. Its advantage depends on the model fitting within 32 GB of VRAM; the Mac’s 256 GB of unified memory can hold models that exceed that limit, making capacity—not benchmark speed—the deciding factor for larger local AI workloads.
Nvidia Would Stop Chip Sales If It Knew AI Was Harming People
Nvidia Chief Executive Jensen Huang said he would stop selling chips to a customer if he knew its AI application was harming people, while cautioning that Nvidia has limited authority over how customers use its general-purpose technology. In an interview with Bloomberg’s Emily Chang, he argued that safety should be built alongside more capable AI systems through containment and monitoring, not treated as a reason to halt development.
ASML’s Capacity Is Strained as Chipmakers Plan Through 2030
ASML’s capacity, rather than a lack of demand, is the constraint as chipmakers plan investments years ahead. Intesa Sanpaolo analyst Antonella Frongillo says equipment makers are discussing demand into 2029 and 2030, though those discussions are not confirmed orders; she sees further upside for ASML despite uncertainty over China sales and the potential impact of the MATCH Act.