OpenAI’s enterprise bounce is squeezing Anthropic’s IPO math: if July share keeps drifting, $190–$200B 2028 revenue assumptions get harder

The newest enterprise market-share datapoint comes from Ramp’s business-user spending data as reported by TechCrunch.

In July 2026, Ramp-based reporting put Anthropic at nearly 44% of enterprise AI spending share, with OpenAI at nearly 40%—a smaller gap than May 2026, when Anthropic was at 41% versus OpenAI at 39%.

Model quality usually moves over quarters. Month-to-month market-share movement inside enterprise spending is more often explained by (1) packaging that reduces friction for IT procurement and (2) service reliability/capacity during peak usage.

Anthropic’s own product/commercial updates point to both: it announced higher Claude Code limits (including doubled five-hour rate limits for Pro, Max, Team, and seat-based Enterprise plans) and removed certain peak-hour reductions, alongside a compute-capacity expansion story via a SpaceX partnership.

Put differently: if Anthropic’s enterprise motion depends on “agents + tool access + higher-commitment plans,” then any competitor counter-positioning from OpenAI—pricing, packaging, or deployment speed—can quickly translate into share shifts because procurement cycles are driven by monthly budget allocation rather than pure research results.

Reuters reports that the valuation work for Anthropic’s IPO is hinging on a $190B–$200B 2028 revenue forecast, using enterprise value-to-revenue multiples applied to those longer-dated numbers.

That structure creates a cliff for any sustained share re-rating: if enterprise spending share doesn’t persist at the level required to justify that 2028 revenue forecast, the multiple math becomes harder to defend—especially for a 2028-anchor that is less typical than near-term (current-year) forecasting.

Agent-like workloads multiply tool calls and longer interaction loops. That increases demand for (1) inference capacity, (2) throughput at peak, and (3) quota discipline so enterprise customers don’t hit “rate/limit walls” that break workflows.

Anthropic’s announced limit increases and compute partnership are therefore not marketing fluff: they’re a direct attempt to keep agent-style enterprise usage within predictable service envelopes. If that works, it supports higher enterprise retention and expansion. If it doesn’t, OpenAI can gain share by offering more predictable enterprise throughput or better-aligned packaging for agents.

A single month’s share drift doesn’t “break” an IPO valuation. What matters is persistence.

You want to see whether the enterprise gap continues narrowing in subsequent monthly Ramp updates, and whether Anthropic’s enterprise limit/capacity initiatives translate into faster re-expansion of enterprise share (or at least stabilization). If OpenAI keeps clawing back, the market may start to discount the probability distribution around the $190B–$200B 2028 revenue range.

Even though OpenAI and Anthropic are the headline competitors, the competitive re-rating can spill into public names tied to the underlying enterprise stack.

Reuters explicitly referenced public-market comps using enterprise value-to-revenue multiples (including Palantir and Cloudflare) to frame how the market might think about scale and growth. If enterprise AI spending becomes more distributed (multi-lab, multi-agent deployments), infrastructure-heavy platforms can benefit from broader adoption and continued high spend.