Version 4

Anthropic has overtaken OpenAI in U.S. business AI adoption — 34.4% vs. 32.3% per Ramp's May 2026 AI Index [1] — driven primarily by Claude Code, which reached $2.5B in annualized revenue by February 2026 [1]. In mid-May, Anthropic executed two simultaneous strategic moves: it acquired SDK and MCP tooling firm Stainless for a reported ~$300M [2][3], whose tools also served OpenAI, Google, Meta, and Stripe, and confirmed that Andrej Karpathy — OpenAI co-founder and former Tesla AI lead — has joined to lead Claude pretraining research and form a new Claude-focused research team [6][7][8]. Meanwhile, Google's A2A (Agent-to-Agent) interoperability protocol has crossed 150 adopting organizations and reached enterprise production use within its first year [13], and both Anthropic and OpenAI are separately building Palantir-style consulting and deployment arms to embed AI into enterprise accounts [12].

Enterprise AI contracts embed into procurement cycles and compliance workflows, making switching costly — so the current land-grab has compounding consequences [1]. The three leading vendors are now pursuing structurally different moats: Anthropic is acquiring the developer toolchain and foundational research talent; OpenAI is partnering with hardware incumbents and deploying human consultants; Google is building open interoperability protocols with meaningful early adoption. The parallel build-out of consulting arms by both Anthropic and OpenAI signals that the battle is now being fought at the enterprise relationship layer, not just the model capability layer.

Through most of 2025, OpenAI held a commanding lead in enterprise AI adoption — Anthropic's share of U.S. business AI spend sat at roughly 8% on Ramp's corporate credit card index in May 2025 [1]. By May 2026, that figure had jumped to 34.4%, crossing OpenAI — now at 32.3% — for the first time. The primary engine of that reversal was Claude Code, Anthropic's autonomous coding agent, which accumulated $2.5 billion in annualized revenue as early as February 2026 [1]. Anthropic's total annualized revenue reached $30 billion by April 2026, surpassing OpenAI's approximately $24 billion [1]. The structural reason this shift may prove durable is the stickiness of enterprise contracts: unlike consumer subscriptions, enterprise AI deployments embed into procurement budgets, internal workflows, and compliance approvals, making switching require a new procurement cycle rather than deleting an app [1].

On May 18–19, Anthropic made two high-signal moves in rapid succession. First, it announced the acquisition of Stainless — the company behind all official Anthropic SDKs since the Claude API launched, and the SDK and MCP tooling provider for OpenAI, Google, Meta, Stripe, and hundreds of other companies — for a reported approximately $300 million [2][3]. The competitive framing has hardened since initial reports: outlets including Digitimes and Forbes now use language of Anthropic actively 'forcing OpenAI and Google to rebuild or migrate SDK tooling' and 'cutting off' competitors' SDK access [4][5], though whether Anthropic will actually restrict access remains unconfirmed. Second, multiple major outlets confirmed that Andrej Karpathy — OpenAI co-founder and former head of Tesla's AI team — has joined Anthropic specifically to lead Claude pretraining research and form a new Claude-focused research team [6][7][8]. The hire is widely read as a landmark talent signal, with community observers framing the combination of Stainless plus Karpathy as deliberate, coordinated positioning [9].

OpenAI's enterprise response has taken a different form. It announced a partnership with Dell on May 18 to bring Codex into hybrid and on-premises enterprise environments, connecting the coding agent to internal codebases, documentation, and operational workflows that enterprises are unwilling to expose to cloud APIs [10]. Some observers characterize this as a defensive moat — a recognition that large enterprises with sensitive data require an on-premises deployment option before they will commit [11]. OpenAI is also expanding Codex beyond software development into broader knowledge work, and has established a $4 billion Deployment Company designed to embed AI engineers directly inside enterprise accounts [1]. Both Anthropic and OpenAI are now explicitly building Palantir-style consulting and deployment arms, targeting private equity and large enterprises with high-touch human engagement [12] — a structural convergence in go-to-market that suggests both companies view the relationship layer, not just the model layer, as the decisive enterprise battleground.

Google has emerged as a visible third competitor. Its A2A (Agent-to-Agent) protocol — announced at Google Cloud Next 2026 — has crossed 150 adopting organizations and reached enterprise production use within its first year [13][14]. The protocol has attracted attention from major ERP ecosystems, with analysts noting it could reshape SAP's AI integration strategy [15], and it has been positioned as an open interoperability standard that positions Google Cloud as the infrastructure layer for multi-vendor enterprise agent deployments [16][17]. The broader agent protocol landscape now includes competing standards — A2A, MCP, ACP, and ANP — creating an emerging standardization battle alongside the model competition [18][19]. The Stainless acquisition directly affects Google as a former Stainless client, adding a new dimension to its competitive calculus with Anthropic [4][5]. The three companies' enterprise strategies diverge sharply: Anthropic is acquiring the developer toolchain and pre-training talent; OpenAI is partnering with hardware incumbents and embedding human consultants; Google is building open interoperability protocols and leveraging existing cloud relationships.