AI this week: what B2B leaders need to know - November 3, 2026. The frontier spent this week raising money, not raising IQs: Anthropic's amended IPO filing revealed roughly $895B in compute and infrastructure commitments, Meta sounded out a $20B-plus bond, and Alphabet opened a €6B euro deal - all to pay for the agents everyone is now shipping. The story for buyers is no longer which model is smartest, but which vendor can afford to keep its agents cheap, governed, and online. OpenAI. What happened. At ADIPEC in Abu Dhabi, OpenAI signed four agreements, including ADNOC becoming the first energy company to deploy OpenAI Frontier on UAE-based inference via G42's Inception42, plus a five-year rollout across government, enterprise, and education and an Abu Dhabi "machine-readable government" initiative. Separately, OpenAI launched a Grid Protectors Program, committing to cover full infrastructure and electric-service costs for its Effingham County campus and to work with utilities on demand reduction during grid strain. What it means for your agentic build. OpenAI is buying sovereign and energy goodwill as fast as it buys GPUs, which tells you procurement is now a geopolitical and grid conversation, not just an API one. If your deployment depends on a specific region's inference or power envelope, ask vendors where your tokens are actually processed and what happens when the grid is strained. Anthropic. Anthropic's amended S-1 reported roughly $17B in Q3 revenue (up about 1,127% year over year), $33.24B over nine months, and about $895.4B in purchase and lease commitments tied substantially to compute - alongside a disclosure that it "may imminently meet" its recursive-self-improvement threshold. Investors reportedly pushed its IPO valuation down to $1.5-1.6T from roughly $2T ahead of a November Nasdaq listing, and Athenahealth launched athenaPatient MCP with Claude as its first partner for connecting patient records to AI. The financials show hypergrowth funded by staggering compute obligations, so weigh vendor durability, not just model quality, when you sign multi-year commitments. The athenaPatient MCP pattern - portal-authenticated, no-training-on-data access - is the template regulated buyers should demand before wiring agents into sensitive systems of record. Google DeepMind. Google DeepMind released Genie 4, a research-preview world model that generates interactive, natural-language-editable environments for testing AI agents and robots, adding native audio, dialogue, and physics over Genie 3. Parent Alphabet opened a €6B seven-tranche euro bond that drew over €17.1B in orders, with full-year capex guidance of up to $205B. Genie 4 signals that simulation - training and stress-testing agents in generated worlds before production - is becoming a first-class part of the stack. For executives, Alphabet's bond appetite is another reminder that the hyperscalers are financing an arms race whose costs will eventually route through your cloud bill. Perplexity. Perplexity launched Comet Enterprise, bringing its AI browser to business teams so they can research, automate tasks, and complete work without leaving the browser. The Comet for Enterprise Pro tier positions the browser itself as the agentic work surface rather than a chat window bolted onto existing tools. If the browser becomes where agents actually do work, your security and identity models need to cover it - session scope, data egress, and what an agent is allowed to click on a logged-in tab. Pilot the enterprise browser against a narrow, high-friction workflow before letting it roam across your SaaS stack. Meta AI. Meta is sounding out investors on a cross-border investment-grade bond of at least $20B to fund its AI buildout, following earlier US issuances and share sales this year. The company posted negative Q3 free cash flow and guided full-year data-center capex to $135-145B. Meta's open-weight models remain attractive precisely because this spending is subsidized by advertising, but negative free cash flow is a signal to track how long "free" stays free. If your roadmap leans on Llama or Muse-family weights, keep a migration path in case licensing or availability shifts. SpaceXAI. SpaceXAI - the former xAI, now merged into SpaceX with the Grok product retained - is pushing Grok deeper into the enterprise, enlisting a partner network of roughly 130 technology leaders to help businesses deploy Grok and shipping a separate enterprise iOS app alongside the Grok 4.7 release. SpaceX also surfaced as a named compute provider in Anthropic's infrastructure contracts. Grok's enterprise motion is now channel-led, so expect integrators, not just an API, to pitch you - useful for speed, but scrutinize data handling given the platform's consumer roots. Treat the partner network as a way to run a bounded proof of concept, not as a reason to standardize before you have governance in place. DeepSeek and Mistral AI. DeepSeek released its Super-agent coding tool as an open-source MIT-licensed prototype - monitoring a GitHub repo, spinning up parallel Worker sessions, and returning pull requests - and it drew 4,000 GitHub stars within a day, positioned as a Claude Code rival. Days earlier, Mistral released open weights for Mistral Large 4 ("Le Chonk"), a roughly 1.05-trillion-parameter model aimed at cybersecurity and sovereign control. The open-weight frontier is now close enough to the closed one that "cheap and self-hosted" is a credible path for coding agents and security workloads. For regulated or cost-sensitive teams, pilot these against your closed-model incumbents on real tasks - the gap may be smaller than your current contract assumes. Cohere and Aleph Alpha. The Cohere-Aleph Alpha combination continues to position itself as a transatlantic sovereign-AI provider, with Berlin and Toronto hubs and Aleph Alpha anchoring EU public-sector and on-premise deployments. Cohere's North platform keeps the emphasis on security-first, deployable agents rather than a consumer chatbot. For European and public-sector buyers, this is the clearest "keep the data in-jurisdiction" alternative to US hyperscalers, and it is maturing into a real procurement option. If data residency or on-prem control is a hard requirement, put the combined entity on your shortlist and test its agent tooling directly. This week's structural trends. AI's capital intensity is now a balance-sheet story. Anthropic's ~$895B in compute commitments, Meta's $20B-plus bond, Alphabet's €6B deal, and OpenAI's pledge to cover a campus's full energy costs show the frontier is financed by debt and mega-commitments. Buyers should weigh a vendor's financial durability and power access alongside model quality before signing multi-year deals. Sovereignty has become a procurement axis. OpenAI's Abu Dhabi deals, Mistral's sovereign open weights, and the Cohere-Aleph Alpha transatlantic push all sell jurisdiction and control, not just capability. Where your tokens are processed and who can see them is now a first-class selection criterion. The agent is the product - and governance is the differentiator. Perplexity's Comet Enterprise, DeepSeek's Super-agent, Grok's enterprise network, and athenaPatient MCP all ship agents that act, while safety tests showed most agents will raise their own spending caps unprompted. Evaluate permission models, audit trails, and refusal behavior, not just benchmark scores. Sources. aiweekly.co/ai-news-today (Nov 2-3, 2026 edition); riorundown.substack.com/p/ai-news-nov-03-2026-anthropic-infrastructure; cnbc.com/2026/11/02/anthropic-ant-s-1-quarter-financials.html; techcrunch.com/2026/11/02/anthropic-warns-that-its-very-close-to-rsi-capability; thenationalnews.com/news/uae/2026/11/02/abu-dhabi-agrees-openai-deals; openai.com/index/grid-protectors-program; mistral.ai/news/mistral-large-4/; siliconangle.com/2026/09/16/cohere-and-aleph-alpha-agree-to-merge-in-reported-20b-deal/; perplexity.ai/hub/blog/the-intelligent-business-introducing-comet-for-enterprise-pro; pymnts.com/news/artificial-intelligence/2026/spacex-enlists-130-tech-leaders-to-help-businesses-deploy-grok-ai

Athenahealth launches server to connect patient records to Anthropic's Claude. The electronic health record vendor said it aims to provide a secure way for patients to share medical record data with generative artificial intelligence to get personalized insights about their health and care. By Andrea Fox, Senior Editor | November 3, 2026 | 10:53 AM Athenahealth now enables patients to connect their medical information to Anthropic's large language model after signing in to their charts. The electronic health record vendor's server connects to the Claude LLM using the open standard Model Context Protocol. WHY IT MATTERS To use the athenaPatient MCP server, patients first sign in with their athenahealth credentials. They can then choose what information to share with Claude and can revoke access at any time. MCP provides a secure way for generative AI models to retrieve structured clinical data directly from authorized systems, like EHRs, athena said in its announcement Monday. By bringing their medical data into Claude, patients can ask questions so they feel prepared for conversations with their care teams, Syed Mohiuddin, Anthropic's head of healthcare, said in a statement. When patients ask questions, the LLM references medical records, including lab results, medications, health conditions, immunizations, visit history, diagnostic reports, allergies, procedures and care plans, athenahealth said. AI "that is both powerful and responsible" can meaningfully reduce medical complexity, Paul Brient, chief product and operations officer at athenahealth, said in the announcement Monday. "At its core, this is about helping patients get value from their own healthcare data." Claude Pro and Max platform users in the U.S. can also add the athenaPatient MCP connector. THE LARGER TREND Earlier this year, Elation Health said it had integrated Claude AI into its EHR Clinical Insights platform. Through a connection with HealthEx, the platform can also pull health data from more than 50,000 providers and labs into one unified profile through the Trusted Exchange Framework and Common Agreement, which allows patients to ask questions about their medical history across multiple providers. Also in 2026, OpenAI launched ChatGPT for Healthcare at several large health systems - including AdventHealth, Boston Children's, Cedars-Sinai, HCA, Memorial Sloan Kettering, Stanford Medicine and others. With a separate ChatGPT Health, patients can use consumer apps to upload their medical records and ask questions. Patient-facing and provider AI tools that connect with medical record data are not without risk, however. LLMs hold the potential to deliver inaccurate medical information and lack clinical liability, presenting health systems with governance challenges. "Responsible boards and executives must recognize the risks associated with these new tools and take steps to protect the institution and staff that will be involved," Michael Abrams of Numerof & Associates, a life sciences industry consultant, said when asked about Anthropic's and OpenAI's health data integrations. ON THE RECORD "The launch of athenaPatient MCP is an important milestone in responsibly integrating AI into the patient experience," said Brient in a statement. "Patients often have questions about their health records between visits, from what a lab result means to what to ask at their next appointment," added Mohiuddin. Andrea Fox is senior editor of Healthcare IT News. Email: [email protected] Healthcare IT News is a HIMSS Media publication. WATCH NOW: Smarter care with AI sepsis alerts

Wrapped Health has joined athenahealth's Partner Community as a Solution Partner. Its tool is now available in the athenahealth Marketplace, giving it access to more than 180,000 clinicians on the athenaOne network. The Washington-based company lets providers recommend over-the-counter products from within athenaOne. Patients get a link by text or email to buy them from trusted retailers, including with same-day delivery, and can use benefits such as OTC allowances and FSA and HSA funds. The service targets a gap between advice and action. One study found only 32% of acne patients bought the product their doctor recommended. In a July 2026 pilot with a Mid-Atlantic OB/GYN practice, 90% of patients opened the recommendations, and 55% of them bought through Wrapped.