AI & Money

OpenAI announced it had submitted a confidential draft S-1 registration statement to the US Securities and Exchange Commission, the first formal step toward a public offering. The filing does not trigger an immediate IPO but signals that OpenAI is preparing the legal and financial infrastructure to go public — and that it intends to do so on terms that reflect its new-for-profit structure following its governance conversion. The timing alongside the SpaceX IPO and a broad tech-capital market rally is notable.OpenAI →

Bloomberg reported that Apollo and Blackstone were working on a roughly $36 billion debt package to fund Anthropic's infrastructure expansion, days after the lab's $65 billion Series H equity round. The structure would route long-duration compute and data-centre spend through private-credit markets rather than equity. The deal underlined that the AI race is now as much about compute financing as model releases, with private capital increasingly underwriting the data-centre build-outs that frontier training depends on.Reuters report

Blackstone announced a joint venture with Google to build a new TPU-based compute-as-a-service cloud, committing approximately $5 billion of equity. The structure parallels the private-equity-backed data-centre and power deals that have characterised AI capex through 2025-2026, applied here to compute capacity rather than to electrons and concrete. The venture is timed with Google I/O 2026 and is intended to broaden TPU availability to third-party customers as Google moves to subscription-style compute pricing on its higher Gemini tiers. The deal is a useful pointer on how AI compute build-outs are being financed: not solely from Big Tech's operating cash flows, but increasingly through joint ventures that move long-duration infrastructure spend off corporate balance sheets and onto private-capital ones.Blackstone press release

OpenAI began testing in-product advertising in ChatGPT, scoped to the Free and Go subscription tiers in supported regions. The company says ads run alongside answers rather than influencing the answers themselves, and that paid tiers remain ad-free. Coverage noted that this is the first time an ad surface has appeared inside the main ChatGPT interface, after years of public statements suggesting the product would stay subscription-only. The move sets up a direct contrast with Anthropic, which has continued to commit publicly to keeping Claude ad-free. For a product class where the LLM increasingly chooses what content a user sees, the question of whether advertisers can influence that ranking — directly or indirectly — becomes a much larger one than for a traditional search engine.OpenAI help-centre post

The Financial Times reported that Anthropic was securing terms for a new funding round at a valuation of around $900 billion — within striking distance of a trillion-dollar private mark and several multiples of where the company stood twelve months earlier. The figure is reporting on in-progress talks rather than a closed round. The valuation, set alongside the company's growing share of business AI spend and its compute deal with xAI, frames Anthropic as the lab that closed its competitive gap with OpenAI faster than nearly anyone forecast a year ago. It also implies a fundraising environment in which only a handful of labs can credibly justify the capex needed to compete on frontier models.Financial Times report

xAI announced that Anthropic will take substantial training and inference capacity on its Colossus supercomputer, framed by the two companies as a way for xAI to monetise underused infrastructure and for Anthropic to relieve its well-documented compute squeeze. Reuters Breakingviews described the deal as commercially logical for both sides despite the two firms being direct rivals on frontier models. The agreement marks the first time a frontier lab has publicly leased significant capacity from another frontier lab, blurring the lines between competitor and supplier in a market where compute scarcity has become the binding constraint on model release schedules.xAI partnership announcement

Nvidia agreed to invest up to $2.1 billion in IREN through a five-year option to buy 30 million shares at $70 each, alongside a separate $3.4 billion AI-cloud agreement under which IREN will deploy Blackwell GPUs. The companies said IREN's Sweetwater, Texas campus could grow from 2 GW of planned capacity to as much as 5 GW of Nvidia-powered infrastructure. IREN — originally a Bitcoin miner now pivoting to AI compute — has rallied 51% YTD on top of 285% in 2025 and already holds a $9.7 billion AI-capacity deal with Microsoft. Critics frame Nvidia's pattern of investing in firms that then buy its chips as “circular”; CEO Jensen Huang has rejected that label. The deal joins recent Nvidia stakes in OpenAI, Marvell, Corning, CoreWeave and Nebius across the AI compute stack.Bloomberg ↗

In Import AI 455, Anthropic co-founder Jack Clark argues a parallel "machine economy" is starting to form inside the existing economy, populated by firms that are capital-heavy (owning compute), opex-heavy (renting AI services) and increasingly labour-light. He expects these companies to trade more and more with one another over time, and eventually for fully autonomous AI-run corporations to emerge — a structural shift he says will surface "profoundly weird" questions on inequality, redistribution and governance. Clark anchors the thesis in capital flows already on the board: Recursive Superintelligence (Rocktäschel, Socher, Clune) raised $500M at a $4B valuation from GV and Nvidia in April 2026 to automate the AI research pipeline; Ineffable Intelligence raised $1.1B at $5.1B; OpenAI has publicly committed to an "automated AI research intern by September 2026"; Anthropic and DeepMind are publishing on automated alignment research. The aggregate is hundreds of billions of capital sunk specifically into firms whose product is automating the production of AI itself.Import AI ↗

Former DeepMind reinforcement-learning lead David Silver, best known for AlphaZero, has raised $1.1 billion at a $5.1 billion valuation for his London-based lab Ineffable Intelligence, only months after founding the company. Sequoia and Lightspeed led the round, with Index, Google, Nvidia, the British Business Bank and the UK's Sovereign AI fund participating. Silver has pledged to donate his personal proceeds to high-impact charities, and the lab is positioned to build a 'superlearner' that acquires capability through reinforcement learning rather than human-generated data. The round is the latest in a pattern of London 'coconut rounds' — billion-dollar early-stage raises by star DeepMind alumni — alongside Yann LeCun's AMI Labs ($1.03 billion) and Tim Rocktäschel's Recursive Superintelligence (around $500 million to $1 billion). Together with Bezos's Project Prometheus, the cluster signals London consolidating as a serious AI hub anchored on DeepMind talent.TechCrunch ↗

AI agents are found to have directly influenced $262 billion in US holiday retail sales, forcing banks and payment networks to confront a structural shift: their systems were designed for human-initiated transactions, not autonomous agentic commerce where AI makes spending decisions and triggers payments on behalf of consumers. Payment processors begin emergency architecture reviews; regulators open inquiries into liability frameworks for AI-initiated transactions as the line between a recommendation and a transaction blurs.Source: Fintech Weekly

Anthropic closed a $30 billion Series G at a $380 billion post-money valuation. It is the second-largest private venture deal in recorded history, behind only OpenAI's $122 billion round closed weeks later. The round formalises a two-horse frontier-lab race in funding terms: OpenAI and Anthropic together account for over $150 billion in private capital raised in a single quarter. The valuation prices Anthropic above all but a handful of public technology companies. It also gives the lab the runway to keep buying frontier-scale compute, datacentre commitments and talent against OpenAI without ceding ground.Crunchbase report

Media analyses and earnings reporting indicate that the largest technology companies are collectively planning capital expenditure of more than $600 to $650 billion in 2026 on AI-related infrastructure, covering data centres, accelerators, and compute capacity. Strong forecasts from ASML and TSMC reinforced the signal that the AI spending boom was intact through the first quarter of 2026, even as questions mounted about utilisation, depreciation schedules, and grid availability. The scale pushes hyperscaler capex well above the combined capital budgets of the US oil majors and utilities sectors for the year, locking in multi-year commitments that shape semiconductor, power, and real-estate markets well beyond 2026.Source: Bloomberg / Reuters