The Enterprise AI Vote Is In: Anthropic Won

Anthropic is reportedly in talks to raise $30 billion at a $900 billion valuation, the same week new Ramp AI Index data showed 34.4% of businesses now pay for Anthropic on at least one workflow versus 32.3% paying for OpenAI. That is the first time in any comparable enterprise benchmark that Anthropic has crossed OpenAI on paid usage. The same Ramp survey put 73% of new enterprise AI purchases in 2026 going to Anthropic. The enterprise market has stopped hedging.

The “two-horse race” framing is over. There’s a horse, and there’s a horse the financial press still writes about because it’s the more interesting story.

If you’re heading into your 2027 planning cycle with a “we use both” answer, the data is telling you something specific about what your peers actually did when the procurement card came out.

Quick Verdict

Signal	What It Says
34.4% of businesses now pay for Anthropic vs 32.3% for OpenAI	First time Anthropic leads in paid enterprise adoption
73% of new enterprise AI purchases in 2026 going to Anthropic	The marginal vendor decision is consolidating around Claude
Anthropic ARR: $30B (April 2026) → $44B+ (May 2026)	47% jump in roughly six weeks. That’s the curve, not the rumor
Gross margins: 38% a year ago → 70% today	Unit economics now look like a software platform, not a research lab
Reported valuation talks: $30B raise at $900B valuation	Would surpass OpenAI’s $852B March 2026 round for the first time
OpenAI valuation: $852B (March 2026)	Higher valuation, lower revenue. One number gets revised next
Your real move this quarter	Decide on platform commitment before the next renewal locks in legacy pricing

What the New Data Actually Says

The 34.4% versus 32.3% number is the part most leaders are misreading because they are still in the headspace where OpenAI is the default and Anthropic is the alternative. The data has flipped that mental model.

Two adoption questions matter for vendor decisions, and they tell different stories. The first is “who uses what model at all.” On that question, OpenAI still has the consumer footprint, more usage at the free tier, and a deeper brand presence. The second is “who pays for it.” Paid usage is the only one that maps to procurement budgets, vendor scorecards, and the renewal conversation you are about to have. Anthropic now leads on that second number.

The 73% new-purchase share is the sharper signal. New purchases are the leading indicator for every enterprise software market that has ever consolidated. When the share of new spend moves above two-thirds toward one vendor, the installed base follows on a two-to-three year lag as renewals come up. That is the exact mechanism that has played out in first-time enterprise AI deals since Q1, and the 73% number is the same trend now confirmed on a wider sample.

The third number is the one CFOs are reading first. Anthropic’s annualized revenue run rate moved from roughly $30 billion in early April to $44 billion-plus by mid-May, per a Semi Analysis report. That is a 47% jump in about six weeks. Public companies do not show that curve. AWS at its fastest did not show that curve. The closest historical comp is the early hyperscaler revenue ramp, and the AI infrastructure market is compressing what took a decade into a single fiscal year.

The Margin Story Is the Real Headline

The revenue number is loud. The margin number is the one that changes the investment case.

Gross margins moved from roughly 38% a year ago to about 70% today, per the same Semi Analysis reporting. A 38% gross margin business is an AI research lab that sells access. A 70% gross margin business is a software platform. That margin band is where Salesforce, ServiceNow, and Snowflake live. It is the band that supports a $900 billion valuation on a non-speculative basis, and it is the band that supports the kind of operating efficiency that public-market investors price.

Three things shift when the margin profile shifts.

The pricing posture changes. A 38% gross margin business has to keep raising prices to grow. A 70% gross margin business can hold prices flat, win on volume, and still post the kind of EBITDA curve that defends an IPO valuation. The pricing risk for buyers compresses. The window of “subsidized pre-IPO pricing” that hammered other AI vendors after their IPOs is narrower for Anthropic than it was for OpenAI heading into its own IPO prep.

The investment case widens. Sovereign funds, pension allocators, and the strategic balance sheets at Google and Amazon allocate against software-platform economics differently than they allocate against research-lab economics. The 70% gross margin number is the line item that unlocks the next tier of capital, and the $30 billion raise being discussed is the first round that prices that unlock.

The competitive moat repositions. Anthropic has spent two years arguing that its constitutional AI and safety training produce a better enterprise product. The margin number argues that the bet has worked twice over. The product wins are real, and the cost-to-serve at scale is lower than the alternative. Two moats are stacking, not one.

OpenAI’s revenue is still growing. The growth rate is the part that compressed. The valuation gap, with OpenAI at $852 billion against the $900 billion Anthropic is reportedly raising at, is the first time a comparison has put Anthropic on top. That gap closes or inverts on the next funding round, on the next set of revenue prints, or on the IPO either company files first.

Why is Anthropic suddenly winning enterprise paid adoption?

Anthropic is now winning enterprise paid adoption because three factors compounded inside the same six-month window: Claude’s frontier model quality moved ahead of the GPT-5 series on the benchmarks enterprises actually evaluate against (coding, agentic tool use, long-context reasoning), the Claude Code product line became the default for developer workflows that translate directly into IT procurement, and the partner network of integrators, hyperscalers, and consulting firms now pushes Claude as the recommended platform inside the same Big 4 engagements that used to default to GPT.

Four mechanisms behind the 34.4% paid adoption lead:

1. Coding workflows became the wedge. Claude leads on the SWE-bench Verified benchmark, and the developer audience for AI coding tools translates directly into enterprise procurement once one team standardizes.
2. Agentic tool use raised the switching cost in Anthropic’s favor. Once an enterprise builds agents on Claude’s tool-use schema, the lock-in is the agent fleet, not the model contract. Migration cost grew faster than model contract value.
3. Hyperscaler distribution closed the access gap. Claude on AWS Bedrock, Google Vertex AI, and Microsoft Foundry means buying Anthropic no longer requires a separate vendor relationship. Procurement friction dropped to near-zero.
4. Consulting firms standardized on Claude Code. The PwC-Anthropic alliance pattern is being replicated by the rest of the Big 4. Implementation partners drive 60%-plus of enterprise vendor decisions, and they have voted with their staffing models.

A buyer doing a clean-sheet vendor evaluation in May 2026 lands on Claude as the default. That is the meaning of “73% of new enterprise spend.” Not preference. Default.

The IPO Window Is the Strategic Variable

Both companies are heading for the IPO window inside the next 18 months. The order matters.

OpenAI’s $852 billion March 2026 round was structured as IPO prep. The same pattern that produced the Sora shutdown and the consumer side-quest cleanup is the operating profile of a company pricing itself for a public listing. The $1 trillion-plus IPO target is real, the cost discipline that hits enterprise pricing post-IPO is also real, and any buyer renewing an OpenAI contract in 2027 is going to be renewing into a different pricing curve than the one they signed.

Anthropic’s reported $30 billion raise at $900 billion is the same pattern, twelve months ahead of schedule. The IPO that priced the AI infrastructure layer will be one of these two companies, and the order of filings will shape the buyer market for the next decade. The company that goes first sets the valuation comp. The company that goes second has to price against the first one’s curve.

Three buyer-side reads on the order question.

The 70% gross margin number suggests Anthropic files first. Bankers price on the cleanest fundamentals available. The Anthropic margin profile is the one that supports a non-controversial S-1. OpenAI’s revenue is larger, the cost-to-serve is heavier, and the path to defending a $1 trillion valuation in public-market scrutiny requires either margin expansion or a narrative about future products that the market may or may not buy.

The vendor renewal cycle inverts. Buyers used to negotiate hardest with the vendor that needed the deal more. Inside an IPO window, both vendors need the disclosed numbers. Negotiating power shifts to whichever buyer has the cleanest exit path and the longest contract horizon. Multi-year commitments at locked pricing are the move. Annual renewals at list price are not.

Lock-in risk goes up before it goes down. The IPO window is the period when each vendor pushes hardest on the enterprise contract structures that produce sticky revenue: minimum commits, custom model fine-tuning, dedicated compute, integration depth. Those structures are the same ones that produce vendor lock-in two years out. The lock-in conversation happens now, not at renewal.

What Consolidation Actually Means for Your Stack

A consolidating market does not mean buyers should pick the winner and ride it. It means buyers should price the consolidation premium and decide whether they want to pay it.

Three concrete shifts inside the next four quarters.

The “two-vendor minimum” rule of thumb is dead. Most enterprise AI policies through 2025 prescribed using at least two model providers to avoid single-vendor risk. That made sense when the model quality was comparable and the integration surfaces were generic. It makes less sense in a market where one vendor has 70%-plus of the new spend, the better unit economics, and the deeper integration into the consulting and hyperscaler layers. A second vendor for redundancy still makes sense. A second vendor for parity does not.

Model-agnostic architecture moves from optional to required. The model-agnostic workflow argument gets stronger, not weaker, in a consolidating market. The point is not to use every vendor equally. The point is to keep the cost of switching one vendor for another inside an acceptable boundary if the consolidation pattern continues. If Anthropic’s reported $900 billion raise materializes, your switching cost from OpenAI to Anthropic is the line item that determines whether your AI program is strategically positioned or strategically stranded.

The platform bet is now a board-level decision. Picking a primary AI vendor was a procurement question through 2025. The capital markets implications, the IPO timing, the partner network depth, and the integration surface area now make it a platform question that ladders to the board. The CIO can run the evaluation. The decision has to be ratified at the level where ten-year strategic bets get made.

My Read on Where This Lands

Three observations on what the data is telling buyers heading into renewal season.

The OpenAI valuation gap closes inside two quarters. The $852 billion versus reported $900 billion comparison is the first inversion. If the Anthropic revenue curve holds at the current slope and OpenAI’s growth continues to compress, the next funding round for either company prices the inversion as durable. The buyers who hedge on the assumption that OpenAI’s brand equity defends its valuation are betting against the operating math that is moving in real time.

The “ChatGPT for consumer, Claude for enterprise” framing oversimplifies what is happening. That framing held in Q1. The May data has Claude pulling ahead on paid usage across the workforce, not just enterprise procurement contracts. The consumer side is still OpenAI’s, but the line between consumer and prosumer is the line that disappears first when employees self-enable on personal tools. The personal-tool spend is moving toward Claude faster than the enterprise procurement spend is moving toward Claude, and the consumer foothold OpenAI has used to defend its position is the foothold most at risk inside twelve months.

The IPO that gets filed first sets the floor for the entire AI infrastructure category. Snowflake’s IPO priced the data cloud category for the rest of the field. ServiceNow’s IPO priced workflow automation. Whichever of these two files first prices AI infrastructure. The current data argues that Anthropic is the more likely first filer on margins, and OpenAI is the more likely first filer on capital appetite. Either order changes how the rest of the category gets capitalized for the next five years.

What This Changes for Your Vendor Bet

Three actions, sized for any enterprise running a meaningful AI line item. Doable inside 90 days. Will position your contract structure ahead of the IPO window for either vendor.

1. Audit your switching cost between Claude and GPT this quarter. Pick the highest-spend AI workflow in your stack and estimate the engineering cost, retraining cost, and contract penalty cost of moving it to the other vendor. The number is your lock-in exposure. If it is large, the renewal conversation is more important than your team is treating it. If it is small, you have more negotiating room than your vendor’s account manager is letting you believe.
2. Rewrite your AI vendor evaluation rubric to include capital markets risk. Add IPO timeline, valuation gap, and gross margin trajectory as scoring categories. They sound like CFO concerns. They are buyer concerns now. A vendor heading into a $1 trillion IPO will reprice your contract after the listing. A vendor with a 70% gross margin profile has less incentive to. The rubric should reflect the difference.
3. Lock multi-year pricing on at least one major workload before the next round closes. If the consolidation pattern holds, the pricing power shifts to whichever vendor wins the next funding round at the higher valuation. Multi-year commits at today’s pricing are the cheapest insurance available against the pricing reset that follows an IPO. The vendors will offer the multi-year terms because the deferred revenue helps their own S-1. The terms are negotiable inside this window in a way they will not be inside the next one.

The platform bet is the one that compounds. The vendor whose partner network, coding tool footprint, and consulting alliances are deepest gets the next cycle’s procurement default by inertia. The 73% new-purchase number is the inertia in motion.

Bottom Line

The enterprise market just voted. Anthropic leads on paid adoption. Anthropic gets the majority of new spend. Anthropic is reportedly raising at a valuation that would surpass OpenAI’s most recent round. The growth curve and the margin curve both moved in Anthropic’s favor inside the same six-week window, and the IPO that prices the entire AI infrastructure category is now likely to be Anthropic’s S-1 as much as it is OpenAI’s.

That does not mean every enterprise should move every workload to Claude. It means every enterprise should stop running a vendor strategy that assumes the two companies are at parity. They are not. The procurement, contract structure, and architecture decisions that get made in the next four quarters land inside a market that already consolidated. Treating the consolidation as still-in-progress is the version of the bet that ages worst.

Audit the switching cost. Rewrite the rubric. Lock the multi-year price on the workload you are most certain about. The IPO window closes the optionality the renewal cycle is still giving you, and the buyers who use the window get a different deal than the ones who wait.

The vote is in. The question is whether your next contract reflects the result.

---

Related Reading:

• ChatGPT Built the Market. Claude Is Winning It.
• Anthropic Out-Earns OpenAI. What That Tells You.
• OpenAI’s IPO Is Coming. Your AI Budget Is Next.
• Big 4 Just Standardized on Claude Code
• Your AI Stack Has an Expiration Date
• Claude Code Is the Business Operating System
• The Anthropic Partner Network Is the SMB Implementation Path
• The AI Enablement Illusion Is Costing You Talent