OpenAI's new revenue chief just fired shots at both its biggest partner and its biggest rival—claiming Anthropic's reported $30 billion revenue is "inflated" by roughly $8 billion due to accounting treatment. For enterprise buyers evaluating AI vendors, this isn't just industry drama. It's a masterclass in how accounting choices can obscure the real competitive landscape—and why you need to look beyond the headlines when picking your AI stack.
The Revenue Claims: What Each Side Says
Anthropic announced $30 billion in annualized revenue in March 2026, up 1,400% year-over-year from $9 billion at the end of 2025. That's a staggering tripling of revenue in 90 days. The company now serves 300,000+ business customers, with over 500 customers spending $1 million or more annually—including 8 of the Fortune 10.
OpenAI hit $25 billion in annualized revenue in February 2026, up from $20 billion at the end of 2025, according to CFO Sarah Friar. The company grew 3x year-over-year ($2B in 2023 → $6B in 2024 → $20B in 2025 → $25B in early 2026) and serves 900 million+ weekly active ChatGPT users.
On the surface, Anthropic appears to have pulled ahead. But OpenAI's revenue chief, in a memo sent to investors, claims Anthropic's $30B figure is inflated by approximately $8 billion due to how it reports revenue from cloud partnerships with Amazon, Google, and Microsoft.
The Accounting Controversy: Gross vs. Net Revenue
Here's where it gets technical—and why CFOs should care.
Anthropic reports revenue from cloud resellers on a gross basis. This means when a customer buys Claude through AWS Bedrock or Google Cloud Vertex AI, Anthropic counts the total end-customer spend as revenue and books the partner payout as an expense. This inflates the top line while maintaining the same bottom-line margin.
OpenAI, by contrast, reports revenue on a net basis for similar partnerships. When a customer uses GPT models through Microsoft Azure OpenAI Service, OpenAI only counts its net share after the partner cut as revenue.
Example: If a customer pays $100 to AWS for Claude API calls, and AWS keeps a 30% distribution fee:
- Anthropic's gross reporting: $100 revenue, $30 cost of revenue = $70 gross profit
- OpenAI's net reporting: $70 revenue, $0 partner cost = $70 gross profit
Same economics. Different top-line numbers.
If OpenAI's $8B adjustment is accurate, Anthropic's revenue on a net basis would be approximately $22 billion—still impressive growth, but no longer ahead of OpenAI's $25 billion.
Why This Matters for Enterprise Buyers
Revenue accounting methods don't change product quality or customer satisfaction. Claude is still Claude. ChatGPT is still ChatGPT. But for enterprise leaders evaluating vendors, these numbers matter in three critical ways:
1. Market Validation and Staying Power
Higher revenue (when apples-to-apples) signals market validation. It means more customers are betting real budgets on the technology, which correlates with product-market fit, feature velocity, and long-term viability. If Anthropic's net revenue is $22B vs. OpenAI's $25B, that's a narrower gap than headlines suggest—and it changes the narrative from "Anthropic overtakes OpenAI" to "neck-and-neck race."
For procurement teams: Don't rely on press releases. Ask vendors for net revenue figures and clarify how they account for cloud partner resales. This matters for assessing vendor stability, especially if you're signing multi-year enterprise agreements.
2. Enterprise vs. Consumer Revenue Mix
The revenue composition tells you where each vendor is investing.
Anthropic's revenue is 80% business customers, with an average revenue per customer exceeding $100,000 annually. Over 500 customers now spend $1 million+ per year. This suggests deep enterprise integration—features like compliance controls, SOC 2 Type II, custom model training, and dedicated support.
OpenAI's revenue is split between enterprise and consumer, with 900 million+ weekly ChatGPT users. While exact breakdowns aren't public, the sheer consumer scale suggests OpenAI monetizes heavily through ChatGPT Plus ($20/month subscriptions) and consumer API usage.
For CIOs and CTOs: Anthropic's enterprise focus means product roadmaps prioritize features you care about—audit logs, role-based access control, on-premises deployment options, and compliance certifications. OpenAI's dual focus means you benefit from consumer-scale innovation but may see slower enterprise-specific feature development.
For CFOs: Anthropic's $1M+ customer cohort (500+ accounts) suggests proven ROI at scale. If you're evaluating a seven-figure AI spend, you want vendors with customers already operating at that level—not beta-testing enterprise features on your dime.
3. Compute Infrastructure and Long-Term Capacity
OpenAI claims it's building toward 30 gigawatts of compute capacity by 2030, compared to Anthropic's estimated 7-8 gigawatts by the end of 2027. This 4x difference matters for buyers planning large-scale deployments.
More compute capacity means:
- Lower latency during peak usage (fewer throttling issues)
- Faster model iteration (more frequent releases and improvements)
- Higher request throughput for enterprise workloads (customer service, code generation, document processing)
Anthropic's counterpoint: The company just signed a 3.5 gigawatt TPU deal with Google and Broadcom extending through 2031, prioritizing efficiency over raw scale. Anthropic claims its models achieve comparable performance with 4x less training compute than competitors—suggesting it's optimizing for cost efficiency, not just capacity.
For engineering leaders: If you're planning to process millions of API calls per day (e.g., customer support automation, real-time recommendations), ask vendors about committed capacity, SLA guarantees, and rate limits. Gross revenue doesn't tell you whether they can handle your scale.
The Enterprise Adoption Signal: "Claude Mania"
Beyond the numbers, there's a qualitative signal enterprise buyers should watch: customer sentiment.
At last week's HumanX conference in San Francisco, Glean CEO Arvind Jain described demand for Anthropic's Claude as "Claude mania," calling it "a religion" among business customers. Air India is using Claude Code to speed up custom software development. Cognizant rolled out Claude tools to 350,000 associates for coding, testing, and DevOps workflows.
What's driving the momentum?
-
Claude Code hit $2.5 billion in annualized revenue in February 2026, with 29 million daily installs on VS Code and 4% of all public GitHub commits. For engineering teams, this means proven adoption at scale—not vaporware.
-
Enterprise context windows and hybrid reasoning enable sustained, cross-session conversations with long documents—critical for legal contract review, financial analysis, and technical documentation.
-
The Model Context Protocol (MCP), Anthropic's open standard for connecting Claude to enterprise systems, reduces integration friction. You can link Claude to proprietary datasets, internal APIs, and software tools without bespoke adapters.
OpenAI's counterpoint: The company's Stargate supercomputer project (estimated at $500 billion) and $200 million Department of Defense contract signal massive institutional backing. If you're in government, defense, or regulated industries, OpenAI's government ties may matter more than Anthropic's enterprise customer count.
What This Means for Your AI Vendor Strategy
If you're a CTO or CIO:
- Don't pick vendors based on revenue headlines. Ask for net revenue, enterprise customer counts, and average contract values. Anthropic's 500+ customers at $1M+ annually is a stronger signal than $30B in gross revenue.
- Evaluate compute capacity against your workload. If you're processing millions of API calls per day, ask about committed capacity, SLA guarantees, and rate limits.
- Test both platforms under your actual use cases. Claude Code's 72.5% on SWE-bench vs. GPT-4's performance on your proprietary codebase may yield different results.
If you're a CFO:
- Model total cost of ownership, not just per-token pricing. Anthropic's efficiency claims (4x less training compute) should translate to lower inference costs at scale—but verify with your engineering team's benchmarks.
- Factor in switching costs. If you're building on one vendor's API, migrating later means retraining embeddings, rewriting prompts, and re-optimizing workflows. Pick the vendor whose roadmap aligns with your 3-year strategy.
If you're a business leader (CMO, COO, CRO):
- Ask your IT team which vendor has better enterprise integrations for your tools. Anthropic's MCP vs. OpenAI's plugin ecosystem may determine how quickly you can deploy AI in Salesforce, Slack, or your CRM.
- Look for proof of ROI in your industry. If you're in aviation (like Air India) or consulting (like Cognizant), Anthropic's case studies are more relevant. If you're in government or defense, OpenAI's DoD contract matters more.
The Real Question: Does Accounting Matter?
For investors and analysts: Yes. Gross vs. net revenue affects valuation multiples, margin analysis, and comparability across companies. OpenAI's criticism of Anthropic's accounting is a fair point for financial modeling.
For enterprise buyers: Not really. Whether Anthropic reports $30B gross or $22B net doesn't change Claude's performance on your use cases, its compliance certifications, or its uptime SLAs.
What does matter:
- Customer concentration risk: If Anthropic's 500+ $1M+ customers include 8 of the Fortune 10, that's strong diversification. If 50% of revenue comes from 3 customers, that's a red flag.
- Burn rate and runway: Both companies are burning billions on compute. OpenAI's $13B from Microsoft and Anthropic's $8B from Amazon provide multi-year runways—but ask about their path to profitability.
- Product velocity: Anthropic shipped Claude Opus 4.6, Claude Sonnet 4.5, and Claude Code in 12 months. OpenAI shipped GPT-4 Turbo, GPT-4o, and o1-preview. Which roadmap aligns with your needs?
Want to calculate your own AI ROI? Try our AI ROI Calculator — takes 60 seconds and shows projected savings, payback period, and 3-year ROI.
Continue Reading
For more on enterprise AI vendor evaluation and deployment strategies:
- How Enterprise AI Leaders Evaluate ROI: Beyond the Hype
- The Hidden Costs of AI Deployments: What CFOs Need to Know
- Vendor Lock-In and the Multi-Model Strategy for Enterprise AI
Sources:
