Anthropic (ANT): The Greatest Business Ever Built? Inside the $40 Trillion Opportunity for Enterprise AI

Key takeaways

• Anthropic filed its S-1 on 26 October 2026 to list on Nasdaq under the ticker ANT. H1 2026 revenue was $16.29bn, up from $1.16bn a year earlier.

• Enterprise spend is compounding. More than 11,900 customers now spend $100K+ a year with Anthropic, up from about 200 at the end of 2024. 1,900+ spend $1M+, and about 250 spend $10M+. Net revenue retention is about 900%.

• The market is enormous and barely touched. Anthropic puts its addressable market at $40 trillion. By one estimate there are about a million businesses worldwide that could spend $100K+ on Claude without noticing, so Anthropic has reached roughly 1% of them.

• Adoption is still shallow. Gallup's tracking puts daily AI use at work at only 10–20%. Microsoft has paid Copilot on about 7% of its commercial Microsoft 365 seats.

• The UK is behind the US, and that is the opportunity. About 72% of Anthropic's revenue still comes from the US. Grant Thornton UK, Barclays and the UK Government show what is possible, but most UK enterprises are only at the start of adoption.

• Boards need a business case they can trust. Only 25% of AI initiatives have delivered their expected ROI (IBM), and only 37% of organisations report any EBIT impact from AI (McKinsey). A board-ready business case is what releases the investment needed to scale.

• Human-led change management is the single biggest unlock. The technology is ready. Value only appears when people change how they work, so the winners will be the enterprises that lead with people, then get data, security and governance right to support them. The ones that buy the most licences won't necessarily win.

What the Anthropic S-1 tells us

I have read a lot of S-1s. This one is different.

Anthropic's filing shows a business that went from $1.16bn in first-half revenue in 2025 to $16.29bn in the first half of 2026. That is roughly 14x growth in a year, at a scale where most companies are happy with 30%. Q2 alone was about $11.6bn.

The IPO is expected to raise around $100bn, which would make it the largest in history, at a valuation widely discussed in the $1.5–2 trillion range. On the latest All-In podcast, David Sacks called it "probably going to be the biggest venture outcome of all time", and Chamath Palihapitiya said it was the most incredible business he had ever seen.

Three numbers matter most for enterprise leaders:

1. 80% of revenue is consumption-based. Companies pay for what they use. Revenue grows when Claude does more work inside the business, not when more licences sit unused.

2. Net revenue retention of about 900%. The average customer that was spending with Anthropic a year ago now spends about nine times as much. For context, the median public cloud software company is around 108%.

3. 48% of revenue comes through AWS, Google Cloud and Azure marketplaces. Enterprises are buying Claude with cloud commitments they already have. Procurement friction is falling fast.

The economics are stronger than the headlines suggested. The reported H1 net loss of $80.7bn is mostly a non-cash charge on convertible notes issued to infrastructure partners. Gross margin is around 85%. Anthropic has raised roughly $110bn but burned only around $9–10bn in total, and it holds about $112bn in cash.

The filing is not without risk. Anthropic has around $417bn in compute commitments, and compute, data centres and power dominate its risk factors. It names OpenAI and Google as direct competitors, with Microsoft as both partner and rival. Investors will debate the valuation. For enterprise buyers the signal is clearer: Claude is now core infrastructure, backed by public-market disclosure, governance and scrutiny.

The real story: a huge market with very little penetration

The headline numbers are big. The more interesting fact is how small they still are compared with the opportunity.

Anthropic sizes its total addressable market at $40 trillion. That is not a software number. It is knowledge work: the analysis, drafting, coding, reviewing and decision support done by hundreds of millions of professionals. As David Friedberg put it on All-In, this may be the first business to realise the old Google idea that "your TAM is commerce".

Now look at how little of it has been captured:

• Customer penetration is about 1%. Friedberg estimates there are 200,000–300,000 businesses in the US, and just over a million globally, for which $100K a year is less than 1% of total spend. Anthropic has 11,900 at that level.

• Spend per customer is under 10% of potential. Only ~250 customers spend more than $10M a year. For a global bank, insurer or professional services firm, that is small against its technology and people budgets.

• Daily use is still low. Gallup's workplace tracking shows only 10–20% of employees use AI at work daily or frequently. Many large enterprises still don't let staff connect Claude to email, knowledge bases, Slack or Notion, mainly because of security concerns.

• Usage grows as costs fall. The cost of a given AI task is falling about 10x a year, and agentic workflows use 50–100x more tokens than simple prompting. Every price cut opens up more use cases.

So the 900% retention figure is not a sign that the market is mature. It shows what happens when a company moves from a pilot to real workflows. The spend curve is driven by depth of adoption, and most enterprises have barely started.

The business case is what unlocks the investment

Chamath Palihapitiya made the most useful point for enterprise leaders. Spend under $1M sits with directors and VPs who can experiment freely. Moving to multi-million-dollar commitments needs a cross-functional decision and a clear ROI or total-cost-of-ownership case.

Only about 1,900 companies have made that case so far. Strip out the AI-native start-ups and he estimates only a couple of hundred Global 2000 companies are there. In his words, "the ROI case is still a little speculative, but the potential is there."

The independent research says the same thing:

• Most AI investment has not yet shown up in the P&L. In the IBM Institute for Business Value's 2025 CEO Study of 2,000 CEOs, only 25% of AI initiatives had delivered their expected ROI, and only 16% had scaled enterprise-wide.

• Profit impact is still rare. McKinsey's State of AI 2026 found that only 37% of organisations attribute any EBIT impact to AI, and around 90% of function-level use cases remain in pilot. Only 6% qualify as AI high performers.

• Pilots stall without a business case. Gartner predicted that at least 30% of generative AI projects would be abandoned after proof of concept by the end of 2025, citing poor data quality, inadequate risk controls, escalating costs and unclear business value. MIT's NANDA initiative reported that 95% of the enterprise generative AI pilots it studied showed no measurable P&L impact within six months.

• The leaders treat AI as a measured, enterprise-wide programme. BCG's 2025 research found only 5% of companies were achieving AI value at scale, and that the biggest roadblocks were organisational, not technical. Its 2026 Applied AI Index shows that share rising to nearly half. The companies creating value run AI as an enterprise-wide programme, invest in governance and the workforce alongside the technology, and measure impact directly in the P&L.

The pattern is clear. Boards are not short of enthusiasm for AI. What they lack is a business case they can trust. Taking full advantage of Claude means funding licences and consumption, data and integration work, security and governance, and a serious change programme. That is a multi-million-pound or multi-million-dollar decision, and it competes with every other priority on the board agenda.

Most enterprises need help to build that case. Internal AI teams tend to talk about models and use cases, while CFOs and boards need cost, value, risk and payback. A board-ready Claude business case should include:

1. A value map by workflow: where Claude changes cost, speed, quality or revenue, with a baseline for each.

2. A realistic cost model: licences, consumption, integration, data, security and change management, including how existing AWS, Azure or Google Cloud commitments can fund it.

3. A phased plan with payback: quick wins that prove value in weeks, building to enterprise-wide scale.

4. Risk and governance: how data, security, regulatory and reputational risks are managed.

5. The people plan: how roles, skills and ways of working will change, and how adoption will be driven and measured.

6. KPIs tracked in the P&L: so the board can see the return and release the next stage of investment.

That is exactly where the opportunity sits. The technology is ready. What most enterprises lack is a credible business case, and the ability to change how their people work so that the case comes true.

Unlocking the $40 trillion opportunity: people first

Every signal in this story points the same way. The constraint is not the model. It is people.

• Licences are not adoption. Microsoft has paid Copilot on about 7% of its commercial seats, and many of those still sit underused. Buying access does not change how people work.

• Usage is shallow. Gallup puts daily AI use at work at 10–20%. Most employees with access are not yet using it as part of how they work.

• ROI follows behaviour change. The ~1,900 companies writing $1M+ cheques are the ones where Claude now runs inside real workflows. That is what the 900% net revenue retention measures: people doing more of their work with Claude, every month.

• The leaders invested in people first. Grant Thornton launched GT Augment, a framework for standards and ways of working, before rolling Claude out firm-wide. Barclays built Claude into the tools its colleagues already use every day.

This is why human-led change management is the single biggest factor in unlocking the opportunity. AI changes how people make decisions, collaborate and are measured. That needs leaders who model the change, managers who redesign workflows with their teams, role-based training, champions in every function, and clear answers to the question every employee asks: what does this mean for my job?

Change management is necessary but not sufficient. In our experience, four other factors decide whether it can succeed:

1. Data readiness. Claude is only as useful as the context it can reach. Clean, well-permissioned access to documents, knowledge bases and systems of record makes the difference between a clever chatbot and a colleague that knows your business.

2. Security and trust. Security concerns are the main reason enterprises keep Claude disconnected from email, Slack and knowledge bases, which is where most of the value sits. Identity, access controls, data residency and audit trails need to be designed in, not bolted on.

3. Governance. Clear policies, risk ownership and human oversight let boards, risk teams and regulators say yes with confidence, so the programme can move faster.

4. Integration and measurement. Claude has to live where work happens, connected through MCP to the platforms teams already use, with baselines and KPIs that prove value in weeks, not quarters.

Get the people side right and these four become accelerators. Get it wrong and even the best model becomes another unused licence.

The Copilot gap: why enterprises are looking again

Many of the enterprises we speak to bought Microsoft 365 Copilot in 2024–25 and are still trying to prove value. Microsoft reported more than 30 million paid Copilot seats in July 2026. That sounds large until you set it against roughly 450 million commercial Microsoft 365 seats: about 7% penetration, two years after launch.

The pattern we see is consistent. Copilot is useful inside Office documents and email. But leaders want AI that can do end-to-end work: reason across complex documents, write and review code, and run multi-step agentic workflows connected to their own systems. That is where Claude, Claude Code and the Model Context Protocol (MCP) are winning.

In our own pipeline across the UK, US and Canada, the enterprise conversations are overwhelmingly about Claude, often alongside Microsoft. We see far less demand to standardise on OpenAI, xAI's Grok or Meta's newer models for regulated, enterprise-wide deployments. Open-source models are being explored, but as the All-In hosts noted, the S-1 numbers show they are not slowing Anthropic's enterprise growth. Trust, safety and data governance are deciding factors for boards, and they favour Anthropic.

UK enterprise adoption of Claude: early, and behind the US

The UK is at an earlier stage of adoption than the US. International revenue was $4.5bn in H1 2026, only 27.8% of the total, so the US still drives most of Anthropic's business. Most UK enterprises we speak to are still in pilots or narrow rollouts.

That gap is the opportunity. Anthropic is investing to close it: it has secured space for up to 800 people in London's Knowledge Quarter, and Bloomberg reports that it plans a UK retail share offering through RetailBook as part of the IPO.

A small number of UK organisations show what scaled adoption looks like:

• Grant Thornton UK is rolling out Claude to all partners and employees across Audit, Tax, Advisory and support teams under GT Augment, its AI governance and ways-of-working framework. It is part of a £500m transformation programme.

• Barclays is scaling Claude across the bank. Its Colleague Knowledge Assistant is used by 16,000+ colleagues supporting more than 20 million UK retail customers. Claude classifies and routes around 120,000 client emails a day in Global Markets. The bank expects 50% of its developers to use Claude Code by the end of 2026.

• The UK Government (DSIT and GDS) is working with Anthropic to build and pilot a Claude-powered agentic assistant for GOV.UK, starting with employment services.

What these organisations have in common is not the technology. It is a human-led approach: leadership sponsorship, clear governance, defined use cases, and investment in people and skills before scale.

The US is further ahead, with Canada following

North American adoption is further along and at much larger scale:

• Cognizant (US) is bringing Claude to up to 350,000 employees and using Claude Code across software engineering, testing and DevOps.

• Deloitte has rolled out Claude to more than 470,000 people globally, one of the largest enterprise AI deployments to date.

• KPMG has integrated Claude into its Digital Gateway platform, available to 276,000+ employees.

• Thomson Reuters (Toronto) has expanded its partnership with Anthropic. Claude connects directly to CoCounsel Legal via MCP, and the next generation of CoCounsel is being built on the Claude Agent SDK.

• The Government of Alberta uses Claude Code to review its systems, find vulnerabilities and fix them.

• Anthropic has committed about CAD $10M in Claude credits to Canadian research institutes including Amii, Mila and Vector, and is looking at significant compute capacity in Canada.

The Big Four and the major integrators are not experimenting. They are rebuilding how they deliver work, and their clients will expect the same.

Five strategic recommendations for enterprise leaders

At Generation Digital we help enterprises in the UK, US and Canada adopt Claude and get measurable value from it. These are the five moves we recommend now.

1. Lead with human-led change management.
Make change management the core of the programme, not a workstream at the end. Secure visible executive sponsorship, involve managers in redesigning their own workflows, run role-based training, build a champions network in every function, and track usage and sentiment. Licences don't create value; changed behaviour does.

2. Build a board-ready business case.
Identify 5–10 high-value workflows where Claude can change cost, speed or quality, such as contract review, credit analysis, client onboarding, software modernisation and reporting. Baseline them, model the full cost including change management, and set KPIs your CFO will recognise. Prove value in weeks, not quarters. This is the evidence that moves the board from departmental experiments to an enterprise commitment, and releases each next stage of investment.

3. Get data, security and governance ready.
Grant Thornton's GT Augment is a good model: a clear framework for standards, risk and ways of working before broad rollout. Pair it with clean, well-permissioned data access and security controls your CISO signs off early. This is what lets you connect Claude to the systems where the value sits.

4. Use the procurement routes you already have.
Nearly half of Anthropic's revenue runs through AWS, Google Cloud and Azure. Check whether existing cloud commitments can fund your Claude deployment. It is often the fastest route through procurement.

5. Plan for agents, not just chat.
The next phase is agentic: Claude connected to your systems through MCP, running multi-step workflows with human oversight. Design your integrations and controls now so you can scale when the business case is proven. Connecting Claude to the platforms your teams already use, such as Asana, Miro, Notion, Canva and Glean, is where adoption becomes habit.

Why this matters now

The Anthropic IPO will put Claude in every board pack in 2027. Your competitors will ask the same question you are asking: how do we get real value from this, quickly and safely?

The answer is not another licence. It is a human-led programme that brings your people with you, on secure, well-governed foundations. The enterprises that move first, with a clear strategy and a partner who has done it before, will build an advantage that compounds, just as Anthropic's own revenue does.

Generation Digital is an award-winning AI transformation consultancy and a certified partner in Anthropic's Claude Partner Network, featured in the FT's UK and European fast-growth lists. We work with enterprises across the UK, US and Canada to design, deploy and drive adoption of Claude, from board-ready business cases, strategy and governance to change management, enablement and agentic workflows.

Planning your Claude rollout or trying to get more from an existing one? Talk to our team →

Frequently asked questions

When is the Anthropic IPO and what is the ticker?

Anthropic PBC publicly filed its S-1 with the SEC on 26 October 2026 to list on Nasdaq under the ticker ANT. Goldman Sachs and Morgan Stanley are lead underwriters. The pricing date had not been set at the time of filing.

How much revenue does Anthropic make?

According to the S-1, Anthropic's revenue for the first half of 2026 was $16.29bn, up from $1.16bn in H1 2025. Full-year 2025 revenue was $4.59bn. About 80% of revenue is consumption-based and 20% is subscriptions.

What is Anthropic's total addressable market?

Anthropic sizes its addressable market for tokens and services at around $40 trillion, which reflects the global value of knowledge work rather than software spend. By customer count it is about 1% penetrated: roughly 11,900 businesses spend $100K+ a year, out of an estimated million that could.

How many enterprises use Claude?

As of 30 September 2026, more than 11,900 customers spent over $100,000 with Anthropic in the previous 12 months. More than 1,900 spent over $1M, and about 250 spent over $10M.

Which UK companies use Claude?

Public examples include Grant Thornton UK (firm-wide rollout under GT Augment), Barclays (Claude Code for developers, a Colleague Knowledge Assistant and email triage) and the UK Government, which is piloting a Claude-powered assistant for GOV.UK.

Is Claude better than Microsoft Copilot for enterprises?

They solve different problems and many enterprises use both. Copilot is embedded in Microsoft 365 apps. Claude is stronger for complex reasoning, coding (Claude Code) and agentic workflows connected to enterprise systems through MCP. The right answer depends on your use cases, data and governance.

What is the biggest success factor for enterprise AI adoption?

Human-led change management. The technology is ready, but value only appears when people change how they work. Executive sponsorship, manager-led workflow redesign, role-based training and champions networks matter most. Data readiness, security, governance and integration are the critical supporting factors.

How do you build a business case for Claude that a board will approve?

Map value by workflow against a baseline, then model the full cost: licences, consumption, integration, data, security and change management. Set out a phased plan with early payback, explain how risks are governed, include the people plan, and agree KPIs that are tracked in the P&L. Research from IBM, McKinsey and BCG shows that most AI investment has not yet reached the P&L, so boards expect this level of rigour before they fund scale.

How should an enterprise start with Claude?

Start with people and a small number of measurable, high-value workflows. Put change management at the centre of the programme, get data access, security and governance ready, use existing cloud commitments for procurement, and invest in enablement before scaling. An experienced Anthropic partner can shorten the path from pilot to value.

Sources

• All-In Podcast, Episode 295: "Anthropic S-1 Shocker, Greatest Business Ever?", October 2026

• Anthropic PBC, Form S-1, SEC, 26 October 2026

• CNBC, Anthropic IPO live updates, 26 October 2026

• Grant Thornton UK, GT Augment launch, June 2026

• Anthropic, Barclays scales Claude, October 2026

• Anthropic, UK Government partnership, January 2026

• Anthropic, Cognizant brings Claude to 350,000 employees

• Thomson Reuters, Expanded partnership with Anthropic, May 2026

• BetaKit, Anthropic commits $10M to Canadian institutions, July 2026

• CNBC, Anthropic London expansion, April 2026

• Bloomberg, Anthropic plans UK share offering via RetailBook, October 2026

• IBM Institute for Business Value, 2025 CEO Study: CEOs double down on AI while navigating enterprise hurdles, May 2025

• McKinsey & Company, The State of AI: Global Survey 2026, August 2026

• BCG, The Widening AI Value Gap: Build for the Future 2025, September 2025

• BCG, Applied AI Index 2026: AI is starting to pay off, September 2026

• Gartner, Gartner predicts 30% of generative AI projects will be abandoned after proof of concept by end of 2025, July 2024

• MIT NANDA, The GenAI Divide: State of AI in Business 2025, July 2025 (reported by Virtualization Review)

• Microsoft, FY26 Q4 earnings release, July 2026