September 16: compute locks in as labs become utilities

Anthropic locks in multiple gigawatts of Google/Broadcom compute as revenue triples

Anthropic has signed a new multi-gigawatt deal with Google and Broadcom for next-generation TPU capacity coming online from 2027, calling it "our most significant compute commitment to date to keep pace with our unprecedented growth."

Why it matters

Anthropic's run-rate revenue has surpassed $30 billion, up from approximately $9 billion at the end of 2025, and the number of customers each spending over $1 million annualized has exceeded 1,000, doubling in less than two months. That growth curve only holds if the compute exists to serve it — which is why an AI lab is now, functionally, also a heavy-infrastructure buyer locking in hardware years ahead of demand.

What this means for you

Anthropic is diversifying its hardware bet — running Claude on AWS Trainium, Google TPUs, and NVIDIA GPUs — while treating chip supply as a strategic asset to be secured multiple years out, not bought on demand.

Finance: Revenue tripling in under a year while committing to multi-gigawatt, multi-year infrastructure spend is the clearest sign yet that frontier AI labs' balance sheets now resemble utilities or cloud providers more than software startups — capital intensity is becoming the moat.

Managers: If your vendor is Claude, this signals capacity is being built for years of growth, not a plateau — plan for continued price/availability stability rather than rationing.

Do this: If your team depends on Claude at scale, no urgent action — but factor continued capacity expansion (not scarcity) into 2027 planning.

Signal 4/5· ImportantSource: Anthropic expands partnership with Google and Broadcom for multiple gigawatts of next-generation compute

Also worth knowing

Policy & risk

OpenAI, Anthropic, Google confirm weeks of joint AI safety talks

OpenAI's global policy chief Chris Lehane confirmed the company has spent weeks coordinating with rivals Anthropic and Google DeepMind on AI safety, following Anthropic CEO Dario Amodei's public call to slow frontier AI development.

Why it matters & what to do

Why it matters

This is a rare admission that fierce competitors think the race itself is a risk worth pausing for. It also lands as the Trump administration pushes an opt-in, keep-pace-with-China approach to AI oversight — meaning industry may end up setting the rules regulators haven't.

What this means for you

If the labs that build these models are quietly worried enough to coordinate, it's worth taking model-risk disclosures and safety evaluations more seriously in your own tool choices, not less.

Managers: Expect vendor contracts and procurement questionnaires to start referencing third-party safety evaluators — factor that into any AI tooling rollout timeline.

Do this: Nothing to do yet — just be aware that industry self-regulation, not federal rules, is currently shaping frontier AI safety standards.

Source: OpenAI, Anthropic, Google have been in talks on AI safety for weeks

Signal 3/5· Pay attention

Tools & tactics

GPT-6 Astra can now work inside old business software without an API

OpenAI's GPT-6 Astra, launched last week, works directly inside everyday business applications by controlling the screen like a person would — no API or integration build required.

Why it matters & what to do

Why it matters

Most AI systems require businesses to prepare their data, redesign workflows, and build custom integrations before they can deliver value. Astra changes that: in ChatGPT Work and Codex, it can write code and work through the same applications people use every day—even when those applications don't have an API, meaning businesses can put AI to work within their existing workflows from day one. That collapses a big chunk of the integration work companies currently pay engineers and consultants to build.

What this means for you

If your job involves wrangling data between systems that don't talk to each other, that manual glue-work is now a target for automation, not a safe niche.

Engineers: Custom API integrations and RPA-style scripting for legacy tools lose value fast — early customers are already using Astra for tasks like optimizing GPUs, spotting discrepancies in financial statements, and producing on-brand deliverables, work that used to need bespoke pipelines.

Managers: Deployment timelines shrink — you no longer need a quarter-long integration project before an AI tool can touch real workflows, so budget and headcount plans built around that lag need revisiting.

Do this: If you or your team maintain custom integrations for legacy enterprise software, pilot Astra's computer-use mode on one of those workflows this quarter to see where it can replace the glue code.

Source: GPT-6 Astra: The next generation in intelligence for work

Signal 4/5· Important

Models & capabilities

Salesforce builds its own reasoning model on Nvidia's open Nemotron, bypassing Claude and GPT

Salesforce unveiled Koa, its first reasoning model, built by post-training Nvidia's open-weight Nemotron rather than paying OpenAI or Anthropic to handle reasoning tasks.

Why it matters & what to do

Why it matters

Koa shows how the enterprise world's needs for AI are diverging from what frontier labs offer, which would rather have enterprises spending millions to upload files, code, prompts, and feedback directly into their models. Salesforce says it always wanted to train its own frontier-grade model but lacked a pretrained base model with clear data provenance until Nemotron.

What this means for you

Salesforce is pitching Koa as an open-weight alternative to closed frontier models, trained for specific work tasks rather than open-ended problem-solving, that has never ingested customer data and so can't leak it.

Engineers: Nvidia says Nemotron's architecture makes inference more token-efficient, hitting what it calls the "trifecta" of sovereign AI, fast time-to-first-token, and efficient reasoning.

Finance: Fine-tuning open models for narrow, high-volume workflows can cut token costs versus routing every task to a frontier model — a lever CFOs will want procurement teams tracking.

Managers: Salesforce isn't cutting ties with the labs entirely — it also launched "Claudeforce," letting companies use Claude as an interface while data stays inside Salesforce's own infrastructure.

Do this: If your company runs heavy, repetitive AI workflows, ask your AI/platform team whether a fine-tuned open model could replace frontier-API calls for those specific tasks.

Source: Salesforce and Nvidia's new reasoning model is everything the AI labs should fear

Signal 4/5· Important

Money & markets

OpenAI in early talks for funding round above $1.2 trillion valuation

OpenAI is holding early talks with investors about a new funding round that would value the company at more than $1.2 trillion ahead of an eventual IPO.

Why it matters & what to do

Why it matters

Any decision to move forward hinges on when OpenAI decides to go public, and the talks were initiated by investors rather than the company. That's a signal that capital is still chasing OpenAI even as skepticism about AI spending grows elsewhere.

What this means for you

A $1.2 trillion-plus valuation would make OpenAI one of the most valuable private companies ever, underscoring how much investor money is still betting on frontier AI's long-term payoff.

Finance: Investor-led talks at this scale suggest institutional money sees more upside ahead of an IPO than in waiting, which could reset comparables across the AI sector.

Do this: Nothing to do yet — just be aware that OpenAI's next funding round, if it closes, will reset the bar for AI valuations heading into any IPO.

Source: OpenAI Weighing Funding Round at Over $1.2 Trillion Valuation

Signal 3/5· Pay attention

One line to sound smart

“Frontier AI labs are now locking in hardware years ahead of demand, building balance sheets that look like utilities, not startups.”

Tool worth a look

Claude is the AI assistant this brief is built with — genuinely useful for drafting, summarizing dense material, and thinking through what a development actually means for you. An honest pick, not a paid link.

Try Claude →

Futureproof Daily is researched and written by AI against our editorial standards — see how we work. Sources are linked on each item. Nothing here is financial, investment, or legal advice.