AI Compute Cloud & Neoclouds

Amazon Web Services Chief AI and Technology Officer Matt Wood said the cost of delivering a given level of AI intelligence is falling by one or two orders of magnitude every three to six months outside the most advanced frontier models, a decline he attributed to improving efficiency as companies learn to operationalize frontier models at lower cost. Speaking to CNBC on Thursday, Wood rejected the idea that more capable AI agents must become less safe, saying developers can surround models with controls, guardrails, security and privacy protections while still improving the underlying technology, and pushed back on the notion that companies must choose between development speed and safety. Wood also defended AWS's $220 billion infrastructure investment, pointing to customer demand and internal business metrics, and said he expects additional computing capacity and greater operating efficiency to bring AI to more businesses and users. Accelerating AWS growth helped push Amazon above $3 trillion in market value for the first time in August 2026, after second-quarter results showed AWS revenue growing at its fastest pace since 2021; the stock surged more than 15% in one session, adding nearly $400 billion in market value, before gaining as much as another 5.3% as it crossed the threshold. TD Cowen analyst John Blackledge said AWS's AI business exited a quarter at roughly a $25 billion annual run rate, its fifth consecutive quarter of accelerating revenue growth, and pointed to CEO Andy Jassy's view that AWS could eventually generate $1 trillion in revenue, compared with the $600 billion long-term opportunity Jassy discussed in his April shareholder letter. Amazon expects AWS to remain capacity constrained through 2026 and 2027 because of AI demand, even as it plans to double AWS capacity by the end of 2027.

Reactor, the developer platform for real-time generative video and world models, announced that NVIDIA and Sapphire Ventures have joined its Series A round, which is led by Lightspeed Venture Partners. The company said the investment comes as demand grows for infrastructure capable of running increasingly powerful world models in real time. Reactor's proprietary inference engine, global GPU cloud and low-latency streaming layer let developers and enterprises run models at 60+ FPS and sub-40ms latency through a unified SDK and API. The platform is already used across media and entertainment, where major Hollywood studios, advertising platforms and video streaming services have active projects, and by leading world model labs in the U.S. and internationally. Reactor was co-founded by CEO Alberto Taiuti and CTO Bryce Schmidtchen, both former technical leads on Apple Vision Pro, and its team includes engineers and researchers from Apple, Meta, Google, Adobe, Replicate and Microsoft.

Nebius Group N.V. has acquired Inferize, an inference optimization company, to strengthen the production inference capabilities of its Nebius Token Factory managed inference platform, with the terms of the transaction not disclosed. The acquisition brings Inferize's technology and systems engineering team into Nebius Token Factory, and the technology is designed to reduce the time required to launch and scale large AI models and make inference workloads more elastic. Inferize's technology focuses on addressing cold starts, the time required for AI models to load before serving a request, which can leave assigned GPUs idle when demand rises and new instances are launched or when model weights are updated during workloads such as reinforcement learning. By reducing this idle GPU time, the technology is designed to let capacity scale more closely with actual usage while supporting higher capacity utilization and improved token economics. Management said Inferize brings technology aimed at accelerating how quickly systems respond to changing customer demand, along with a team experienced in GPU systems, and that the team's contribution is expected to extend beyond the initial integration. The deal adds another layer to Nebius' production inference efforts, following Eigen AI's model, kernel- and system-level optimization work and the addition of Clarifai's core team and licensed technology for system-level inference and compute orchestration.

Firmus Grid plans to allocate about half of the shares in its initial public offering to existing shareholders, potentially allowing Nvidia and Blackstone to increase their stakes, Bloomberg News reported on Monday, citing people familiar with the matter. The Australian data centre operator has received investor indications well above the offer size and is seeking to raise as much as A$5.5 billion, or $3.8 billion, including a greenshoe option. The bookbuilding deadline has been brought forward to Thursday from Friday. Firmus priced the IPO at A$11 a share, implying a valuation of about A$43.7 billion, or $30.3 billion, a listing that could rank among Australia's largest IPOs, comparable with Medibank's 2014 offering, which raised just under $5 billion. Firmus has secured commitments from investors including Nvidia and Blackstone as it expands data-centre capacity to meet demand for artificial intelligence computing, with proceeds helping fund GPUs for its first data centre in Batam, Indonesia.

Jim Cramer said on the September 28 episode of Mad Money that Space Exploration Technologies Corp. is developing into a broader technology and infrastructure company, with Starlink, AI computing, and Starship as multiple potential growth engines. Cramer pointed to SpaceX's GPU infrastructure business, saying Elon Musk is renting GPUs bought from NVIDIA to Google at an $11 billion a year run rate and has a similar deal with Anthropic at a $15 billion run rate starting in October, roughly $2.17 billion a month in compute rentals. The company's second-quarter results showed AI revenue of $2.56 billion, up 247% year over year, alongside $14.1 billion in contracted Cloud Services Agreements, though that AI revenue came with a $1.26 billion operating loss. Cramer also highlighted the September 28 Starship flight, which reached orbit and deployed 26 Starlink V3 satellites, but a Raptor Vacuum upper-stage engine shut down prematurely, cutting the planned 10-hour mission to about three hours. SpaceX generated $7.81 billion of second-quarter revenue while spending $18.37 billion on capital expenditures, $15.83 billion of it on AI, and Cramer separately warned that a large amount of restricted stock will become eligible to enter the market over the next year. Insider Monkey reported that 119 hedge fund portfolios held SpaceX at the end of the second quarter of 2026, and the stock trades at a price-to-sales multiple of 85.5x.

Amazon has established an $8 billion special purpose vehicle to buy and lease back Nvidia Grace Blackwell chips to AWS clients, an off balance sheet financing tool for AI hardware as cloud capital demands rise. Morgan Stanley has reinstated Nvidia as a Top Pick, citing its view on AI data center demand and capacity expansion. The Amazon vehicle and the Morgan Stanley call both sit inside a wider Nvidia AI infrastructure story, alongside Nvidia's own US$500 billion ecosystem financing plans. The structure lets AWS keep building AI capacity without stacking all the hardware on its own balance sheet, pointing to new funding structures that can support large orders as AI factories become more capital intensive across cloud and sovereign projects. The clearest early signal on whether the model scales would be other hyperscalers or sovereign AI buyers setting up similar chip vehicles that explicitly name Nvidia hardware, with disclosed sizes and timelines.

Nscale has hired Justin Osofsky, a long-term executive at Meta Platforms, to be its chief operating officer to drive rapid expansion at the AI infrastructure company ahead of its upcoming IPO, Bloomberg News reported, citing a person familiar with the matter. Osofsky, who served Meta for 18 years in various capacities and was most recently its chief partnerships officer, will report to Nscale CEO Josh Payne and lead the company's global operations. A slate of Meta veterans, including former Chief Operating Officer Sheryl Sandberg, already serve Nscale, a neo-cloud operator that rents out computing power to AI developers; Sandberg, who sits on the company's board, helped hire Nick Clegg, the Facebook operator's former head of global affairs, and Fidji Simo, a former OpenAI executive. In September, the London-based data center operator filed for a U.S. IPO, an offering expected to raise as much as $3B, Bloomberg previously reported. Meta announced the resignation in an internal post, with Chief Operating Officer Javier Olivan noting that Osofsky had brought sound judgment and steady leadership that will outlast his time here.

Alphabet has unveiled Argon, the flagship of its Gemini 4 generation, months later than promised after the company scrapped Gemini 3.5 Pro, which Sundar Pichai had said would arrive in June. Google says Argon matches OpenAI's Astra and Anthropic's Opus on key coding and cybersecurity tests, though its own results show it trailing on two of the four coding tests included, and the model has no public release date, going only to select cybersecurity partners under the Trump administration's voluntary pre-release access process. Google is competing on price rather than raw capability, pricing Argon at $2 per million input tokens and $10 per million output tokens with cached input tokens discounted 95%, which Jefferies analyst Brent Thill calls a particularly important competitive move at about half the cost of some rival models. The delay came amid leadership churn, with DeepMind founder and chief executive Demis Hassabis stepping aside and several Gemini leaders leaving while Anthropic and OpenAI kept releasing new top models. The infrastructure side is stronger: Google Cloud revenue grew 82% year over year to $24.8 billion in Q2, customers have lined up a $514 billion backlog, cloud now makes up a little more than 20% of Alphabet's revenue, and its cloud market share has climbed to 14% from 12% at the end of 2025, while Alphabet projects 2026 data center capital spending of $195 billion to $205 billion.

UBS said it sees the artificial intelligence infrastructure market staying hot, keeping CoreWeave in focus on Friday. In a note to clients, UBS analyst Karl Keirstead wrote that the per-hour price of hosted Nvidia GPUs is rising, that the environment is inflationary, and that revenue-per-gigawatt figures are moving higher, now a key pillar of the hyperscaler and neo-cloud bull case. Keirstead noted that CoreWeave disclosed this week it raised per-hour GPU pricing across all SKUs by 25% in July 2026 and has since raised them by another 10% in just the last two to three months. He also flagged the risk that local community and state and local government pushback could throttle the pace of AI data center additions, though he said checks suggested the industry will manage through the hurdle. Keirstead added that the availability and cost of capital could slow the AI buildout, with some projects expected to fade, but said better-positioned players including CoreWeave will not have an issue.

Anthropic's draft IPO prospectus shows roughly $110 billion owed to Amazon Web Services over about the next decade, part of $518 billion in total cloud, computing and infrastructure obligations the AI company has planned. Amazon's share traces back to a pledge of more than $100 billion over 10 years and, like most of that total, is owed even if Anthropic's usage falls short, giving Amazon revenue visibility that ordinary pay-as-you-go cloud usage does not. Spread evenly, the AWS portion runs above $10 billion a year, while Anthropic reportedly spent $7.33 billion on all of its suppliers combined in 2025. Amazon has also invested $18 billion in Anthropic through convertible notes and nonvoting preferred stock, with up to $15 billion more available, and those holdings are now carried at about $190 billion, up from about $61 billion at the end of 2025. Second-quarter net income more than tripled to $62.6 billion, with $53.4 billion of pre-tax, non-operating income coming mostly from the Anthropic investments. Separately, Synopsys and AWS announced a multi-year deal worth more than $1 billion under which Amazon's cloud unit will license chip design blueprints, and AWS now offers Claude Sonnet 5.5 through Amazon Bedrock.

Anthropic's confidential IPO filing shows the AI company's 2025 revenues jumped nearly 12-fold to about $4.6 billion, while operating losses more than doubled to more than $8 billion. Nearly 47% of that revenue, or about $2.16 billion, came through the cloud marketplaces of Amazon and Alphabet's Google, up from 11% in 2023 and 32% in 2024, and Anthropic paid $351 million in distribution fees, equal to 16 cents for every dollar of marketplace revenue. Most revenue, roughly $3.8 billion, came from usage-based Claude customers, with subscription revenue at $789 million, and two unnamed customers each contributed 12% of 2025 revenues. Anthropic's non-cancellable hosting and computing commitments rose from $54.6 billion at the end of 2025 to more than $417 billion by early 2026, covering 3.5 GW of capacity, with a fixed compute commitment of $1.25 billion a month through May 2029. The company raised its 2028 revenue forecast to $190-$200 billion, raised $65 billion at a $965 billion valuation in May, and could post its first quarterly GAAP operating profit in Q3 2026 despite an estimated $10-$15 billion in cumulative losses.

CoreWeave announced availability of the NVIDIA Vera Rubin NVL72, with Cognition as its first production customer, alongside new support for the NVIDIA Vera CPU. Cognition, which uses CoreWeave for training, reinforcement learning and inference, reported that the Vera Rubin NVL72 delivered up to 4.8x higher total token throughput for SWE-2 inference workloads versus a GB200 NVL72 baseline and 3.8x higher output-token throughput for reinforcement-learning workloads. CoreWeave said a Vera rack can contain 128 CPUs and 11,264 cores, theoretically supporting more than 11,000 concurrent isolated environments, and that testing showed more than three times faster agent sandbox startup times compared with an x86 CPU. The company will offer Vera on bare metal using the same operating model and economics as the rest of its infrastructure, aiming to monetize CPU-intensive infrastructure alongside accelerator hours. CoreWeave remains heavily dependent on NVIDIA's technology roadmap and faces competition from hyperscalers and specialized GPU clouds including Microsoft Azure and Nebius Group N.V., which closed four deals in the quarter averaging more than $1 billion each and plans roughly £1.7 billion in U.K. AI compute expansion expected to deliver 65 MW when fully operational in 2027.

Nebius Group has acquired inference optimization startup Inferize and signed new long-term AI data center capacity deals, moves that aim to boost GPU utilization on its Token Factory platform and deepen its role in global AI infrastructure. The company's recent inclusion in the FTSE All-World Index came alongside fresh analyst coverage and large customer commitments from partners including Meta Platforms and Palantir. Nebius Group's narrative projects $29.7 billion in revenue and $2.1 billion in earnings by 2029, requiring 179.9% yearly revenue growth and about a $2.1 billion increase in earnings from $42.4 million today. Some of the most optimistic analysts were already modeling Nebius to reach about US$48.5 billion in revenue and US$1.1 billion in earnings by 2029, well above consensus. The Inferize deal directly targets Nebius's cost base and unit economics by cutting the idle GPU tax on Token Factory, though financing needs and potential dilution remain key risks.

On the evening of September 30, ST Dou Shen (300010) disclosed that its wholly owned subsidiary Dou Shen Hong Kong recently signed a Cloud Computing Service Agreement with a certain customer, with a total contract value of about 119 million US dollars, equivalent to about 800 million yuan. The service period of the agreement is 60 months, and the customer will prepay 40 percent of the total order amount as agreed and pay service fees monthly. ST Dou Shen did not disclose specific customer information, stating that the contract involves confidentiality clauses and that it has completed internal procedures for exemption from disclosure of confidential information. The company said the contract is a large-scale computing power service project for the implementation of its intelligent computing service business, an important practice in cultivating a second growth curve, and reflects the extension of its business from AI applications to the computing power infrastructure field. Wind data shows that since mid-August 2026, ST Dou Shen's stock price has risen by more than 120 percent cumulatively.

ChronoScale Holdings Corporation announced a contract extension with an existing AI infrastructure customer and a separate agreement with a new AI infrastructure customer, moves the company says will lift its annualized revenue run-rate to $1 billion by calendar Q3 2027. The two agreements, combined with ChronoScale's existing customer contracts and current deployment schedule, expand the company's contracted deployment pipeline and overall capacity. Chief Executive Officer Cenly Chen called the expansion and the new customer an important validation of demand and of the company's ability to grow alongside its customers, adding that the company now has agreements in place providing for $1 billion in annualized revenue run rate during calendar 2027. ChronoScale also completed the previously disclosed sale of its Ekso Bionics business unit, a divestiture it says will let it focus operations, management resources and capital allocation on its core accelerated compute and AI infrastructure business. The company, which trades on Nasdaq under the symbol CHRN, was formed through the combination of Applied Digital's cloud business and EKSO Bionics Holdings, Inc. Achievement and timing of the contracted revenue remain subject to risks including timely deployment and commissioning of planned infrastructure capacity, availability of power and equipment, and access to capital on acceptable terms.

PaleBlueDot AI has closed a $200 million Series C financing led by ComputeCore at a valuation of $3.2 billion. The round follows the Silicon Valley company's $150 million Series B announced in January, with existing shareholder B Capital participating alongside other global investors. Founded in 2024 and headquartered in Palo Alto, PaleBlueDot AI operates a Super Intelligence Infrastructure Platform combining self-owned GPU clusters, a GPU marketplace and serverless inference, and its B300 cluster in Japan recently earned NVIDIA Exemplar Cloud status. As of the end of September 2026, the company had signed over $5 billion in customer contracts, with customers from the U.S. and Japan together accounting for more than half of monthly revenue. Proceeds from the Series C will fund additional Super Intelligence compute capacity, and CEO Stephen Watts said the company will broaden its customer base with a focus on frontier labs, Neolabs and enterprises in the U.S. while investing in its full-stack and go-to-market teams.

Oracle shares reversed an early gain and turned lower Thursday after the Financial Times reported that Tencent Holdings had locked in a five-year lease spanning several Oracle data centers across Southeast Asia. Under the reported terms, Tencent would gain access to about 100,000 cutting-edge AI chips it otherwise can't obtain inside China, with the arrangement said to be worth roughly $7 billion and close to a third of that due upfront. Neither company responded when asked to confirm the details. Oracle stock had jumped nearly 2% before Thursday's opening bell on the report, but the shares gave up that advance amid overall market weakness and were down 0.82% at $136.18 at the time of publication. The deal reflects a broader scramble among Chinese tech firms for overseas computing capacity as domestic chip supply stays tight and U.S. export restrictions persist.

La Rosa Holdings Corp. announced it acquired latest-generation NVIDIA B300 GPUs and leased those assets back under a long-term lease expected to generate recurring, long-term cash flows, advancing the company's previously announced transition into the artificial intelligence infrastructure sector. As part of that broader strategic transition, the company said it intends to evaluate strategic alternatives for its real estate operations, including a potential divestiture of its existing real estate businesses, which management believes could further accelerate the shift toward AI infrastructure. La Rosa also announced that Nicholas Adler has been appointed Chief Executive Officer and interim Chief Financial Officer effective October 1, 2026; Adler has served as Chairman of the Board since December 2025, while founder and former CEO Joe La Rosa has stepped down to lead the real estate business and will remain a member of the Board of Directors. The company said it intends to pursue additional opportunities including strategic partnerships and investments to expand its presence in the AI ecosystem, and may also pursue opportunities in alternative high-growth sectors. La Rosa added that it intends to change its corporate name and Nasdaq ticker, with further updates to come as the Board determines appropriate.

CoreWeave, Inc. CRWV is expanding its AI cloud ecosystem through the launch of the CoreWeave Partner Network and a new customer deployment with Ennoble Care. The Partner Network brings together technology integrations validated on CoreWeave Cloud under production conditions, including collaborations with VAST Data, CrowdStrike, Reflection and ClickHouse, among others, covering infrastructure, data services, software, security, models and inference. The network is also expanding into tools used by AI agents, starting with search, with Exa, Parallel Web Systems and You.com joining to give agents built on CoreWeave access to a shared search layer through a single integration. Separately, Ennoble Care selected CoreWeave to support clinical AI inference workloads across its home-based care network, deploying NVIDIA RTX PRO 6000 Blackwell Server Edition nodes through CoreWeave Kubernetes Service with reserved GPU capacity, managed Kubernetes, hardware-enforced isolation, zero-trust controls and direct access to CoreWeave specialists. CoreWeave stated that CKS provides eight to 10 times faster container spin-up times than a general-purpose cloud alternative, and highlighted its MLPerf inference and training results along with three consecutive Platinum rankings in SemiAnalysis ClusterMAX.

Singapore-founded neocloud Aolani announced a partnership with Karman, formerly known as Utilidata, to squeeze more AI compute out of its existing power supply through advanced power orchestration. Under the deal, Aolani will integrate the Karman platform, a dynamic power orchestration system built on a custom NVIDIA Jetson Orin Nano that combines high-resolution metrology with local processing and AI, into its infrastructure stack to measure and dynamically allocate available power across high-density GPU workloads and manage rack-level energy consumption in real time. The joint proof of concept will be deployed on the NVIDIA Blackwell Platform. Karman's platform has already unlocked 33% more compute capacity from existing power infrastructure at its first commercial deployment in North America, potentially translating into a $20M increase in revenue per 1 MW, and the company says it increases tokens per watt by 50% by unlocking stranded power. Aolani CEO Nicholas Chia said the partnership improves the economics of AI workloads and lets customers scale without waiting on the grid, while Karman CEO Josh Brumberger said the two will demonstrate how the platform can fully utilize power already available to data center operators and cloud providers.

CoreWeave unveiled Forge, a unified AI development environment, at its Fully Connected AI cloud conference. The company also introduced the CoreWeave Partner Network, featuring validated technology integrations aimed at enterprise and advanced AI users, and expanded its compute lineup by adding the NVIDIA Vera CPU to support emerging AI agent workloads. CoreWeave runs cloud infrastructure for AI developers and enterprises, and the launch sequence signals how the US-based provider is trying to pull more of the AI build process onto its own stack. The company said the clearest test of the strategy will be future disclosures tying Forge and the Partner Network to measurable backlog or workload share, such as updated figures on how much of its contracted AI capacity is running through Forge-driven workflows in quarterly updates through 2027.

GMI Cloud has raised $668 million in new financing, comprising $223 million in equity for its Series B round and a $445 million credit facility led by CTBC. The Series B was led by ARCHIV, a San Francisco investment firm specializing in AI and robotics, with participation from NVIDIA and Asia-Pacific investors including DSC Investment, Trend Micro, KB Investment, Kyobo Life and KT Corporation. The growth capital will fund capacity expansion in the United States, Taiwan and the rest of APAC, building on the company's Taiwan AI Factory and its Japan sovereign AI initiative, and will also support its inference services and strategic hiring. GMI Cloud said its contracted ARR has reached more than 9x its year-end 2025 level, while live ARR in production has grown more than 4.5x over the same period, and its inference platform now processes approximately 4 trillion tokens per week. Customers include Fireworks, Higgsfield, Nous Research, OpenRouter, Reflection, Cartesia, Trend Micro and Utopai Studios.

Anthropic reported preliminary Q2 2026 revenue of $11.6 billion, overtaking OpenAI's $6.7 billion for the first time and posting its first quarter of positive adjusted operating profit. The figure is up from $787 million a year earlier, a 14-fold year-over-year surge driven by enterprise adoption of Claude Code, and more than doubles the $4.73 billion Anthropic booked in Q1. It also amounts to 2.5 times Anthropic's entire 2025 annual revenue of $4.59 billion, putting the company on an annualized run rate of roughly $46 billion. The result reframes the S-1 filed two days ago with a $2 trillion-plus valuation target, which laid out $518 billion in compute commitments, including $111.1 billion to Google, $110 billion to Amazon and $161.2 billion to Broadcom, with roughly 80% binding and non-cancelable. OpenAI's operating loss, inclusive of stock-based compensation, widened to $12.3 billion in Q2 from $9.3 billion in Q1, though CFO Sarah Friar told employees that July annualized recurring revenue already exceeded the full Q2 total. The shift comes as Anthropic faces regulatory pressure on multiple fronts, including a September 25 D.C. Circuit classification as a supply chain risk under FASCSSA Section 4713 and a September 30 FTC probe into frontier labs. All figures are preliminary and subject to revision; neither company has filed audited public financial statements, and Anthropic's exact adjusted operating profit has not been disclosed.

Hewlett Packard Enterprise has secured a $1.2 billion deal with Vultr to deploy AMD's Helios rack system across AI data centers, the first Helios deal the company has ever announced. HPE CFO Marie Myers said the company also raised its long-term revenue forecasts, lifting its revenue guide for 2027 and 2028, and now projects low to high 50s percent growth in data center networking revenue through 2029. Myers said HPE has always used silicon from multiple vendors and that the AMD partnership is one of many, with the Helios portfolio expected to become an important part of the business. She pointed to explosive data center growth, citing last quarter's deal with Oracle and momentum in self-driving networks, where ServiceNow set a goal of being 100 percent autonomous with its network by 2028.

Cerebras Systems is carrying $25.4 billion in remaining performance obligations as of June 30, 2026, work customers have signed for but the company has not yet delivered. That backlog sits alongside management's raised August forecast for core revenue of $880 million to $890 million for all of 2026, and a 2027 target for core revenue to more than triple. Core revenue in the second quarter of 2026 was $209.9 million, up 103% from a year earlier, most of it tied to the company's cloud service as its OpenAI deployment ramped up. The signed work includes no business from AWS or any other hyperscaler, though Cerebras expects its offering on AWS' Bedrock platform to be generally available in the first quarter of 2027. Management has secured more than 600 megawatts of data center capacity, live now or due by the end of 2027, and said data center space is the industry's bottleneck; core gross margin should hit its low point in the third quarter of 2026 before a significant fourth-quarter improvement. The shares trade 37.3% below their 52-week high at 51.1 times sales, against 3.1 times for the S&P 500.

CoreWeave Inc. announced it will expand its compute portfolio with the NVIDIA Vera CPU, the first CPU designed for AI agents, at its Fully Connected AI cloud conference in San Francisco. The company said rack-scale Vera at CoreWeave puts 128 CPUs and 11,264 cores in a single rack, with BlueField-4 DPUs and Spectrum-X Ethernet switching, providing room for more than 11,000 concurrent agent environments. In testing, CoreWeave achieved more than 3x faster agent sandbox startup times on NVIDIA Vera CPU compared to an x86 CPU. Chen Goldberg, executive vice president of product and engineering at CoreWeave, said general-purpose infrastructure bottlenecks agentic AI and that Vera is the first CPU explicitly designed to accelerate it. CoreWeave said NVIDIA Vera runs bare metal under the same platform, consumption models and economics as the rest of its fleet, and is natively enabled through products such as CoreWeave Sandboxes.

Nvidia has reportedly held talks with insurance companies about shifting risk tied to loans backed by its AI chips, as CEO Jensen Huang pushes to make GPU infrastructure easier to finance beyond Big Tech buyers. According to a Financial Times report on Tuesday, one structure under discussion would insure lenders if smaller cloud operators, or neoclouds, default and the Nvidia chips pledged as collateral cannot be resold for enough to cover the loan. The talks remain preliminary and may not produce deals, and Nvidia has also shared chip-depreciation and future compute-price data with at least one insurer. The effort extends Nvidia's push to turn compute into an investable asset class; in August the company partnered with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR on platforms designed to mobilize more than $500 billion of third-party capital for AI infrastructure. Nvidia's filings say access to capital can constrain less-capitalized AI clouds and model makers, and according to an SEC filing from July 26, Nvidia had $36 billion in typically six-year cloud-service commitments to help selected AI-cloud partners deploy more infrastructure, while reserving the option to provide limited residual-value support on individual projects.

TD Cowen initiated coverage of SpaceX with a Buy rating and a $200 price target on Tuesday, betting that leasing AI data center capacity will quickly become the company's largest business. Analyst John Blackledge expects AI compute leasing to generate $14.4 billion in revenue this year, about 35% of SpaceX's total, before jumping to $66 billion in 2027, or about 58% of sales, making leasing the majority of SpaceX revenue by the first quarter of 2027. SpaceX has four customers leasing its compute capacity: Anthropic, Google, Reflection AI, and an unnamed fourth customer that agreed to a deal starting in December at $1.1 billion per month. TD Cowen expects SpaceX to lease another 4 gigawatts in 2027 and 2028 at $46 billion to $50 billion per gigawatt, and sees terrestrial capacity climbing from 2.1 gigawatts at the end of this year to 6 gigawatts in 2027 and 22 gigawatts by 2031. The bank forecasts 62% annual revenue growth from 2026 through 2031, with EBITDA reaching $297 billion in 2031, and does not expect SpaceX to turn free cash flow positive until 2030.

Anthropic's S-1 prospectus, reported by Reuters on September 28, 2026, outlines a company seeking a valuation exceeding $2 trillion while carrying $518 billion in planned cloud, compute, and infrastructure obligations. That total includes roughly $111.1 billion to Google, $110 billion to Amazon, $31.4 billion to Microsoft, and $161.2 billion to Broadcom, with approximately 80% of the obligations binding and non-cancelable. Revenue reached approximately $4.59 billion in 2025, meaning that for every dollar Anthropic earned it committed roughly $113 to future infrastructure. The filing also details a Founder LLC structure giving seven co-founders, including CEO Dario Amodei, President and Board Chair Daniela Amodei, Chief Compute Officer Tom Brown, and Chris Olah, 50.1% voting power through a Class F share, with the co-founders pledging 80% of their equity to charity and the Long-Term Benefit Trust overseen by figures including former Federal Reserve Chair Ben Bernanke and national security expert Richard Fontaine. On September 25, the D.C. Circuit Court of Appeals upheld a Pentagon classification of Anthropic as a supply chain risk under FASCSSA Section 4713, and reports of NVIDIA considering a $10 billion anchor investment remain unconfirmed.

OpenAI's annualized revenue run rate has reached nearly $70B, up more than 70% since the start of the third quarter, Axios reported on Tuesday, citing people familiar with the figures. OpenAI did not immediately respond to Seeking Alpha's request for comment. Oracle shares jumped about 5% after the Axios report, and Microsoft shares rose to $512; Oracle is a major cloud infrastructure provider to AI companies, including OpenAI. Anthropic's annualized revenue reached about $65B in July, and its revenue rose twelvefold to nearly $4.6B in 2025, according to an IPO prospectus reviewed by Reuters. OpenAI's business-to-business revenue has more than doubled over the same period, while its consumer business added more revenue during the third quarter than it did throughout 2025, according to the report.