AI Infrastructure and Compute
The hardware and cloud stack behind AI, including GPUs, accelerators, data centers, networking, inference chips, and compute supply constraints.
AI Costs Pressure Margins as Google Search Risks Sending Less Traffic
John Coogan’s and Jordi Hays’s discussion on TBPN frames AI as both a business opportunity and a source of pressure on costs and distribution. Coogan said rising memory and AI expenses could weigh on Apple’s margins, while the hosts questioned whether Grok can turn its place on X into a viable route to workplace users. They also treated Google’s AI search as a potential threat to the referrals other sites depend on, while leaving the scale and consequences of these shifts uncertain.
Investors Question AI Spending as Google’s Search Economics Shift
Big Tech’s selloff reflects more than investor unease about rising AI bills, Alex Kantrowitz and Ranjan Roy argue: it exposes uncertainty over whether companies’ established businesses can sustain the spending. Google’s search advertising remains large and profitable, but AI answers could weaken the flow of web traffic that has supported its economics. Roy also says Anthropic’s filing leaves investors without a clear view of its compute costs, while falling model prices raise questions about how durable AI companies’ returns will be.
AI Could Raise Energy Demand While Helping Transform the Energy Industry
OpenAI Chief Executive Sam Altman said AI could reshape the economics of energy in two directions: by increasing demand for power and by giving the energy industry new ways to work. Speaking at the ADIPEC conference in Abu Dhabi, he said the AI industry could “repay the favor” after years of discussing what it needs from energy. His remarks came as OpenAI signed deals to deploy its technology with energy companies and other organizations in the United Arab Emirates.
Anthropic Revenue Jumps Twelvefold as Compute Commitments Reach $895 Billion
Anthropic reported $16.96 billion in third-quarter revenue, nearly twelve times its total a year earlier, alongside a net loss of $23.3 billion. Bloomberg’s Bailey Lipschultz said convertible-note accounting accounted for a substantial part of the loss, while quarterly compute costs were close to $10 billion. Bloomberg Intelligence’s Mandeep Singh said the company’s large future compute commitments reflect a business that must keep investing in infrastructure as it scales.
Anthropic’s Growth Case Hinges on Enterprise Returns and Compute Capacity
Anthropic’s S-1 presents a company growing at extraordinary speed, but the All-In hosts disagree on what its figures establish. David Sacks argues that rising margins, falling model costs and a large potential customer base support a historic business case; Chamath Palihapitiya says the unresolved questions are whether enterprise trials turn into durable, multimillion-dollar contracts and how the costly infrastructure buildout will be financed.
Google Cloud Backlog Grows as Search Ads Face an AI Monetization Problem
John Coogan argues that Alphabet’s results show Google Cloud accelerating while Search advertising growth slows and AI answers remain harder to monetize. The article then turns to two separate subjects: Jordi Hays recounts allegations from a Times report that a Chinese intelligence asset had developed a close relationship with Nvidia CEO Jensen Huang, and the hosts describe trading firms charging recurring fees for simulated challenges that may offer customers a chance at capital.
Microsoft’s $25 Billion Bond Offering Signals More AI Spending Ahead
Microsoft is seeking to raise at least $25 billion in its first broad U.S. bond offering since 2017, a move Bloomberg Intelligence senior credit analyst Robert Schiffman says reinforces his view that AI spending will continue to grow. He argues Microsoft’s strong credit standing gives it room to borrow while continuing to invest and return cash to shareholders. The deal, in his assessment, is another sign that the broader hyperscaler spending cycle has not yet peaked.
Marvell Weighs an $83 Billion Lumentum Deal as Optics Supply Tightens
Bloomberg’s Ryan Gould reported that Marvell Technology is in advanced talks to acquire Lumentum Holdings, a maker of optical components for data centers, in a deal that could make Lumentum a substantial part of Marvell. Gould’s case for the fit is that optics are becoming essential to moving signals between servers as data-center demand grows, and Marvell has said it wants to expand in the field. But with Lumentum valued at about $83 billion and Marvell holding less than $4 billion in cash, how the companies could finance a deal remains unresolved.
AI Compute Demand Is Pushing OpenAI Toward Google’s TPUs
John Coogan and Jordi Hays argue that growing AI compute demand is pushing companies to use a wider range of chips, cloud providers and data-center operators, even as Nvidia remains central. They cite reports that OpenAI plans to buy Google TPUs as evidence that compute needs can outweigh competitive history, while stressing that the market is still early and dependence on Nvidia remains substantial.
Microsoft’s Backlog Grows Beyond OpenAI as Azure Revenue Climbs 48%
Microsoft’s results showed Azure revenue up 48%, while Bloomberg Intelligence analyst Anurag Rana said investors should focus on whether cloud demand is broadening beyond major AI companies. He also said they would look for clues about Microsoft’s capital spending plans, including whether the company might raise its 2027 target.
The US and China Are Now Six Months Apart in AI
Eric Schmidt says the US and China are now about six months apart in AI, narrowing a gap he once put at one or two years. In a live Masters of Scale Summit interview, the former Google chief argues that China has made ground through cheaper models, abundant power and faster deployment, while the US retains advantages in frontier performance, talent and capital. His scorecard ties the two countries, but Schmidt says the contest need not be zero-sum—and warns that US restrictions on foreign researchers risk weakening its own position.
GPU Prices Are Rising Despite a Wave of New Compute Supply
Gavin Baker argues that the AI infrastructure buildout depends on more than expected GPU supply: companies must secure increasingly expensive compute and bring it online quickly enough to earn a return. Delays compound the cost of tight supply by leaving high-value equipment idle, while public listings by Anthropic and SpaceX could, in his view, unlock buying power for private-market investment. He reads Anthropic’s S-1 as evidence that enterprise adoption remains early, reinforcing the case for further investment.
Anduril and Qualcomm Aim to Build a US Laptop Supply Chain
Anduril Industries and Qualcomm are developing BlueChip, a laptop designed and assembled in the United States, as an effort to build a domestic computer supply chain. Bloomberg’s Ed Ludlow reported that the companies see dependence on overseas production, particularly China, as a national-security vulnerability; the project has so far produced a prototype, not a supply chain operating at scale. The laptop is expected to be expensive initially, with public-sector buyers as its first market and cost competitiveness with overseas-made products a longer-term goal.
Anduril and Qualcomm Target a Fully Domestic Laptop Supply Chain
Anduril founder Palmer Luckey argues that the US needs a domestic computer supply chain because military and critical infrastructure systems still rely on machines made in China. In an interview with Bloomberg’s Ed Ludlow, he said Anduril and Qualcomm are building BlueChip to address that dependence, drawing on the defense company’s manufacturing experience. The prototypes are not yet fully US-made: Luckey said they are about 90% US and allied production, with a roadmap to reach 100%.
Google’s AI Advantage Depends on Coordinating Products, Compute, and Research
Google’s AI advantage may depend less on having frontier models than on making its products work as one, argues Margins’ Ranjan Roy, while Big Technology’s Alex Kantrowitz sees model capability as central to building a useful unified assistant. Their discussion also examines whether Google should direct scarce computing capacity to its fast-growing cloud business or AI research, as investors show signs of fatigue with AI fundraising. Cognition CISO Alex Stamos warns that the PHP vulnerability is part of a broader surge in cyberattacks, with forgotten systems and slow patching leaving organizations exposed.
Anthropic’s S-1 Shows $67 Billion Revenue Run Rate and Rising Margins
Anthropic’s S-1, as described by TBPN hosts John Coogan and Jordi Hays, shows a fast-growing company whose improving margins and falling compute costs are set against hundreds of billions of dollars in long-term capacity commitments. The hosts argue that the commitments are less alarming when viewed across their multiyear terms, but say the filing leaves unresolved questions about demand, the sources of revenue tied to terms-of-service violations and the founders’ plan to retain 50.1% of voting power after an IPO.
Anthropic’s IPO Could Deliver Menlo Ventures’ Biggest Return
Menlo Ventures partner Matt Murphy says the firm’s investment in Anthropic is poised to deliver its largest return ever if the AI company goes public. Menlo has backed Anthropic in every round since 2023, and Murphy argues that its S-1 offers investors a rare chance to assess extraordinary growth directly. He also says the competitive test is shifting beyond model quality to include access to compute and how efficiently companies use it.
Meta Seeks Up to $95 Billion in Stock Sales to Fund AI
Meta is seeking to raise as much as $95 billion through a stock offering to fund its AI investments, Bloomberg’s Sarah Frier reports. The plan combines common stock with shares from a new non-voting class, which Meta says could help attract different kinds of investors while preserving some of Mark Zuckerberg’s stake from dilution. Frier said the target exceeds Google’s $85 billion offering earlier this year, though she noted more skepticism around Meta’s plan.
AI Factories Are Being Built to Run Agents, Not Just Models
NVIDIA CEO Jensen Huang argues that AI infrastructure is shifting from systems built to run models to AI factories designed to run agents, which he describes as the basis of a new industrial economy. In his GTC Berlin keynote, he extended that case to physical AI, presenting shared platforms and reference systems for vehicles and robots as part of the effort to build Europe’s AI economy.
AI Infrastructure May Pay Off Before Its Applications Are Clear
Blackstone chief executive Stephen Schwarzman says the firm’s early AI investments grew out of a broader strategy: use information from its businesses to spot economic shifts before they appear in public data, then invest in the infrastructure those shifts require. He points to Blackstone’s $10 billion purchase of a data-center company in 2021 as a bet on AI’s power and computing needs, while arguing that the commercial returns may take years to emerge.
Venture-Backed Startups Account for Most AI Spending, Raising Questions About Durable Demand
Chamath Palihapitiya argues that venture-backed startups account for most AI spending, making current demand a fragile measure of broad adoption; Jason Calacanis counters that startups’ willingness to buy reflects the technology’s practical value. Palo Alto Networks CEO Nikesh Arora warns that AI can help uncover vulnerabilities and accelerate attacks on unpatched systems, exposing the risks of years of deferred maintenance. In France, the hosts disagree over whether unrest and fiscal strain stem chiefly from excessive state spending, disappointing public services, immigration or weak growth.
Xona Raises $250 Million to Build a Commercial GPS Alternative
Xona says GPS’s limitations are becoming more consequential as autonomous vehicles, drones, robotics and other systems rely on precise location with less human oversight. CEO and co-founder Brian Manning argues that smaller, lower-cost satellites in a closer orbit can improve navigation while working with existing GPS devices through software updates. The company has raised $250 million at a $2.45 billion valuation to build and launch its network after receiving FCC approval to operate alongside GPS.
GPU Capacity Still Trades Through Private Quotes and Bespoke Contracts
Compute capacity still trades largely through private quotes, brokers and bilateral contracts, Affaan Mustafa of Jump Trading says, with larger commitments negotiated out of public view. He argues that falling token prices have not necessarily lowered compute costs: cheaper inference can drive more usage, while chips, power, data-center space and capital shape the price of capacity. The financing gap is sharpest for mid-size companies, whose demand may grow unevenly while suppliers need long-term commitments and upfront payments can consume scarce runway.
AI-RAN Puts Network Functions and AI Services on Shared Infrastructure
Nokia and NVIDIA argue that telecom operators can turn network infrastructure into a shared platform for radio access and AI workloads, creating a path from programmable 5G toward 6G. Their AI-RAN collaboration pairs Nokia software with NVIDIA accelerated computing; the companies say it can improve network performance, support AI services at the network edge and help automate network operations.