Big 5 AI Vendor Roundup for the Week of May 4, 2026

May 11, 2026

Research By: Mark Tauschek, Bill Wong, Info-Tech Research Group

Anthropic took center stage this week with its “The Briefing: Financial Services” livestream event and a flurry of distribution and compute announcements. The event was somewhat unusual and not your typical product announcement hype. Anthropic also signed a $1.8 billion Akamai deal on Thursday and a SpaceX/Colossus compute deal earlier in the week and announced a $1.5 billion services joint venture (JV) with Goldman Sachs, Blackstone, and Hellman & Friedman (H&F). OpenAI matched the JV within minutes with a $10 billion private equity (PE) vehicle. Microsoft expanded Copilot Cowork. AWS shipped AgentCore optimization. Google held fire for I/O on May 19.

The model layer is becoming commoditized, so the next arms race is in distribution and compute capacity.

Anthropic: A Different Kind of Vendor Event and a Compute Land Grab

The May 5 event “Briefing: Financial Services” in lower Manhattan broke the mold for vendor keynotes. This is covered in detail in this research note, but here are some of the highlights:

• Dual-message risk disclosure. Two weeks earlier, Anthropic disclosed Mythos's autonomous zero-day capability, held the release, and committed $100 million to defenders via Project Glasswing. Treasury, the Fed, the Bank of England, FCA, NCSC, and Bank of Canada convened emergency briefings. On stage Tuesday, Dario Amodei called it a “moment of danger” and Jamie Dimon said it was “transitory.” Both messages are defensible but the message to customers is obviously tempered from the initial messaging to the cybersecurity community.

• Dimon’s labor impact concerns seemed to shift. At Davos and in March congressional testimony, Dimon warned AI may move “faster than society can absorb.” On Anthropic’s stage, he pointed to NAFTA trade adjustment assistance as the redeployment model, then conceded in the same breath it didn’t work.

• Ten prebuilt finance agents, full M365 integration, Moody’s MCP app. Pitchbook builder, model builder, valuation reviewer, GL reconciler, month-end closer, statement auditor, KYC screener, credit memos, insurance claims, and market researcher agents are all on Claude Opus 4.7 (Vals AI Finance Agent benchmark of 64.37%). Excel, PowerPoint, and Word add-ins become generally available, while Outlook is in beta, and context will carry across all four. New connectors were announced for Verisk, Third Bridge, Fiscal AI, D&B, Experian, GLG, Guidepoint, and IBISWorld. The FIS Financial Crimes Agent on Claude is live at BMO and Amalgamated Bank. The agentic deployment model (managed agents combined with vertical templates and curated connectors) is the template Anthropic will repeat by vertical.

• >80x revenue growth vs. 10x internal projection. Amodei said, “the cone is even wider than I thought,” when announcing the revenue trajectory. Annualized revenue went from ~$9 billion at the end of 2025 to over $30 billion by this month. Dimon, somewhat surprisingly, endorsed the trillion-dollar CapEx buildout. It doesn’t resolve the circular financing question, but rapid demand growth is a positive sign.

• $1.5 billion enterprise AI services JV with Blackstone, Hellman & Friedman, and Goldman Sachs. Anthropic, Blackstone, and H&F are committing $300 million each, with Apollo, General Atlantic, GIC, Leonard Green, and Sequoia backing the initiative as well. Anthropic will be deploying onsite engineers, aimed at mid-market and PE portfolio companies.

• A compute land grab on three fronts. Three deals in eight business days tells you everything you need to know about how tight supply is.

○ Google: up to $40 billion (April 24). $10 billion in cash now at a $350 billion valuation and an additional $30 billion contingent on hitting performance milestones. Consider that along with the previously announced 5-gigawatt TPU commitment with Google and Broadcom. This is another good example of circular financing in the cycle: Google invests $10 billion, Anthropic spends it on Google compute, Google books it as Cloud revenue, rinse and repeat. Google’s Cloud constraint problem (acknowledged on the Q1 earnings call) plus Anthropic’s seemingly insatiable compute hunger makes both sides willing to overlook the fact that Gemini and Claude compete directly.

○ SpaceX Colossus 1 data center lease deal includes over 300 megawatts of compute and over 220,000 NVIDIA GPUs and will be online within the month. That will provide some relief for the compute strain–induced rate-limit pressure that’s been building since January. Pro and Max five-hour subscriber limits were doubled the same week. There is an interesting political subplot with this agreement as Anthropic is fighting a Pentagon contracting dispute over military-use guardrails while signing with Musk, who is suing OpenAI and is the closest tech executive to the current administration.

○ Akamai and Anthropic sign a $1.8 billion content delivery network (CDN) deal that was announced May 7, making it the largest contract in Akamai’s 28-year history. Akamai stock was up 29.6% on the news. The CDN is becoming the AI delivery cloud as Akamai is deploying thousands of NVIDIA RTX PRO 6000 GPUs on a distributed edge platform for inference. Revenue from the deal starts in Q4 2026.

○ Stacked on existing commitments with Amazon (up to 5 gigawatts of Trainium compute and $25 billion in equity), Microsoft/NVIDIA ($30 billion in Azure compute), and Fluidstack ($50 billion investment in compute), Anthropic is scaling compute across a diversified base of every major silicon vendor, three hyperscalers, xAI’s Colossus, and a CDN. Capacity, not model capability, has become the constraint.

OpenAI: $10 Billion DeployCo, GPT-5.5 Instant, Ads Manager

• The Deployment Company (DeployCo) is a Delaware LLC with a $10 billion valuation and $4 billion raised from 19 investors (including TPG, Brookfield, Bain, Advent, SoftBank, and Dragoneer). OpenAI is committing up to $1.5 billion ($500 million now and an optional $1 billion). OpenAI has guaranteed a 17.5% annual return to PE backers over five years, which is unusual for an operating-company JV. Contrast this with Anthropic’s common equity structure plus 80x organic growth, and you see Anthropic letting customers pull while OpenAI is paying investors to push. Both JVs are reportedly acquiring services firms for forward-deployed engineering capacity.

• GPT-5.5 Instant replaces GPT-5.3 Instant as the ChatGPT default. It is touting reduced hallucination in law, medicine, finance, as well as “fewer unnecessary follow-up questions, [reduced] clutter like over formatting and gratuitous emojis.” That statement is an explicit acknowledgment of the AI-tell complaints from users.

• ChatGPT Ads Manager beta and CPC bidding. In direct contrast with Anthropic’s ad-free positioning, IT leaders will need to decide how this might reflect in enterprise AI policy before standardizing.

• GPT-5.5-Cyber will be made available for vetted defenders via expanded Trusted Access program. This is OpenAI’s Mythos counter, except it’s been made more accessible rather than strictly confined to a small group in early preview. There are different arguments on whether broad or restricted model access is safer.

• A Trusted Contact safety feature was added on personal accounts only (not Business, Enterprise, or Edu) following a series of wrongful death and safety lawsuits against the company.

Microsoft: Copilot Cowork Expansion

• Copilot Cowork added inbox orchestration, multi-step research, structured document generation, and web page building (available in the early access Frontier program). It will be available for $99/user/month in the M365 E7 bundle with general availability expected later this month.

• The 2026 Work Trend Index finds that 49% of Copilot conversations support cognitive work and 58% of AI users say they’re producing work they couldn't a year ago. Take it for what it is, which is marketing-grade research, not independent benchmarking.

• The strategic question: Microsoft is competing with both Anthropic and OpenAI inside its own Office apps, so you might expect aggressive M365 E7 pricing in the third quarter and beyond. Members renewing Copilot Business or E5 should run hard comparisons against Claude on M365 and ChatGPT from the newly formed DeployCo before signing multiyear agreements.

Google: Holding Fire for I/O Conference (Mostly)

The I/O opening keynote is on May 19, while the pre-I/O Android Show is May 12. Look for announcements on Gemini 4 and the Remy 24/7 agentic assistant.

• The week’s biggest Google story was the $40B Anthropic investment (covered in the Anthropic section). Doubling down on a direct competitor to Gemini tells you Google would rather lock up Anthropic’s compute spend than win the model race outright.

• Workspace Intelligence rollout continues more widely in May with real-time Gemini context across Gmail, Chat, Calendar, and Drive. This is Google’s answer to Anthropic’s M365 add-ins.

• Google AI Ultra Lite tier is in development (codename Neon) and will come in packages ranging between $20 Pro and $250 Ultra. Anthropic and OpenAI already have subscriptions priced at $100, so Google is filling the gap.

Amazon: AgentCore Optimization, Quick in the Wild

• Bedrock AgentCore optimization (preview) includes recommendations, batch evaluations, and A/B tests for agent performance. The observe, evaluate, improve loop for production agents will be necessary for any organization deploying at scale, or agent quality regression would likely go undetected.

• Amazon Q Developer end-of-support. The functionality will be consolidating around Quick (desktop assistant) and Bedrock-based developer tooling. Quick preview is widely available, so keep an eye out for users installing it and possibly running afoul of policy.

• OpenAI on Bedrock (GPT-5.5, GPT-5.4, Codex, Managed Agents) remains in limited preview. With both OpenAI and Anthropic shipping vertical agents through their own JVs, AWS becomes a neutral platform for everyone's agents.

Our Take

Distribution and compute, not models, were the big stories of the week. Anthropic ($1.5 billion JV, 80x growth, common equity) and OpenAI ($10 billion JV, 17.5% guaranteed PE returns) are both buying their way into enterprise workflows because organic enterprise sales can’t keep pace with the CapEx curve.

The compute story is bubbling up with Anthropic closing deals with Google ($40 billion), SpaceX (300 megawatts of compute), and Akamai ($1.8 billion in CDN services) in eight business days, on top of existing commitments from Amazon, Microsoft, NVIDIA, and Broadcom. Every major hyperscaler is now both invested in and competing with Anthropic. That is either the strongest demand signal of the cycle or the clearest illustration of circular AI financing we’ve seen yet. Probably both. Demand is getting real, but capacity is the binding constraint.