An internal memo from OpenAI’s Chief Revenue Officer Denise Dresser, leaked by The Verge on April 13, 2026, has exposed the company’s full Q2 battle plan — a new AI model codenamed “Spud”, an enterprise agent platform called “Frontier”, and some of the bluntest public criticism yet of rival Anthropic. The document was intended for OpenAI employees, not the press. It is now one of the most-discussed leaks in the AI industry this year.
The memo makes five core priorities clear. First, Spud — described by Dresser as “an important step in the intelligence foundation for the next generation of work” — is being positioned not just as a smarter chatbot, but as the engine beneath an entire enterprise super app.
OpenAI CEO Sam Altman completed Spud’s pre-training on March 24, 2026, shut down the consumer video tool Sora entirely to redirect compute, and told employees it is “a few weeks” from public release. OpenAI President Greg Brockman has called it “two years of research” with a “big model feel.”
Second, OpenAI is building an agent platform called “Frontier”, which Dresser writes is designed to become “the default platform for enterprise agents”. The goal, in her own words: move from being a product vendor to being operating infrastructure.
Every workflow that runs through Frontier makes OpenAI harder to replace — that is not an accident, it is the strategy. A deployment engine called “DeployCo” will help large enterprises roll it all out at scale.
Third, OpenAI’s recently announced Amazon partnership is expanding far beyond model access. The memo describes an “Amazon Stateful Runtime Environment” that gives AI agents persistent memory and continuity across complex, multi-step business processes. Since the deal’s announcement in late February, demand has been, in Dresser’s words, “frankly staggering”.
The memo’s sharpest section, however, is a direct assault on Anthropic. Dresser accuses the company of building its brand on “fear, restriction, and the idea that a small group of elites should control AI”. More specifically, she claims Anthropic’s widely reported $30 billion revenue run rate is overstated by roughly $8 billion — because the company counts gross revenue from its deals with Google and Amazon, rather than the net figure OpenAI uses. If the accusation holds, Anthropic’s real run rate would sit at around $22 billion, placing it behind OpenAI’s reported $24–25 billion. Neither claim can be independently verified — both companies are still private.
What This Means for India
For a country of 1.4 billion people building its digital economy on IT services, this matters enormously. Companies like TCS, Infosys, and Wipro have built entire service arms around integrating different AI tools for enterprise clients. If OpenAI successfully bundles a closed, full-stack platform — models, agents, deployment, memory — it could reduce the need for third-party integrators entirely.
Indian AI startups face a harder choice: build on OpenAI’s expanding ecosystem and risk deep vendor lock-in, or pivot to open-source alternatives for independence. At a time when India is racing to lead in AI, the push for BharatGen and sovereign AI infrastructure becomes more urgent than ever — because platforms like Frontier are designed for deep, long-term enterprise integration with no easy exit.
Key Details at a Glance
- Leaked memo author: Denise Dresser, Chief Revenue Officer, OpenAI — dated April 2026, reported by The Verge
- “Spud” model: Pre-training completed March 24, 2026; expected release April–May 2026; may launch as GPT-5.5 or GPT-6
- Frontier platform: OpenAI’s enterprise agent infrastructure, designed to replace third-party AI integration layers
- Amazon deal: Includes a “Stateful Runtime Environment” giving AI agents persistent memory across business workflows
- Anthropic revenue claim: OpenAI alleges Anthropic overstates its run rate by ~$8B; neither figure is independently verifiable as both companies are private
What Happens Next
Spud’s public release is expected sometime between April and May 2026 — Altman has confirmed it is weeks away, though no official date has been set. Anthropic has not responded publicly to the revenue allegations. Both companies are preparing for IPOs later in 2026, and the accounting dispute will face far greater scrutiny once either firm files public disclosures.
FAQs
What is OpenAI’s “Spud” model?
“Spud” is the internal codename for OpenAI’s next major AI model, which completed pre-training on March 24, 2026. It is expected to launch as either GPT-5.5 or GPT-6 and is designed to serve as the foundation for OpenAI’s enterprise super app and autonomous agent ecosystem.
Did OpenAI really say Anthropic is inflating its revenue?
Yes — according to the leaked memo, OpenAI’s CRO Denise Dresser claims Anthropic overstates its $30 billion revenue run rate by roughly $8 billion due to how it accounts for revenue-sharing deals with Google and Amazon. However, neither company is publicly traded, so these figures cannot be independently verified.
How does this OpenAI vs Anthropic battle affect Indian companies?
Indian IT services giants and AI startups are caught in the middle. If OpenAI’s Frontier platform becomes the default enterprise AI infrastructure, it could reduce demand for third-party integrators — a core revenue stream for Indian IT firms. Startups building on OpenAI’s ecosystem also risk vendor lock-in, making a strong case for India’s own sovereign AI initiatives.