SNI: WEEK 18

Welcome to all the AI news that matters this week – across tech, biopharma, medtech, advanced manufacturing and insurance. The wins, the fails and the somewhere in-betweens.

tl;dr: we can see clearly now

With so much moving so quickly, only the brave or naive call something 'a trend' in the universe of AI. A data series - describing adoption, competitive positioning or value - may appear robust one week. Only for it to be revealed as an ephemeral wisp or blip the next.

But this week, fortune favoured the brave. Dozens of hypotheses became trends. Wisps and blips turned to concrete. Solid. Clearer. More substantive.

For instance, it is now entirely obvious that we all need to enjoy subsidised AI use while its lasts. The bill for those tokens we're burning will soon become due.

GitHub Copilot is moving to consumption-based billing on 1 June. And commentators nodded, acknowledging that metering was always inevitable. With more providers expected to follow suit soon.

Which meant DeepSeek's V4 release, priced at a fraction of Western closed-source rivals had even greater significance. The largesse of 'frontier models for everything' will wane more quickly given viable alternatives.

The trend of Anthropic winning also solidified. After last week's revelation that the secondary market was already valuing it more highly, TechCrunch reported Anthropic could close a $900bn-plus valuation round within the next two weeks, pricing the company a sliver above OpenAI's $852bn. Proactive Investors and Gizmodo both pointed out this is eclipsing OpenAI as the most valuable private AI company.

Two days earlier, Google had publicly committed up to $40bn to Anthropic across cloud, equity and TPU access. It also cemented Google's position against Nvidia. The deal locks Google's silicon inside a frontier-lab customer, even though Anthropic competes with its Gemini stack.

The Microsoft–OpenAI partnership, meanwhile, continued to reprice its mutual value in the opposite direction. OpenAI's models will now also be hosted on Amazon Web Services and Bedrock - ending Microsoft Azure's cloud exclusivity. The Register read it as OpenAI 'jumping out of Microsoft's bed and into Amazon's Bedrock'; the LA Times described it as the end of one of the most consequential exclusivity arrangements in cloud computing.

But it wasn't all back-pedalling. Accenture's announcement that it would roll Copilot out to all 743,000 employees cemented improving expectations around Microsoft's oft-maligned tool.

The trend towards regulation also became firmer. Brussels ordered Google to open up Android to rival assistants. OpenEvidence, meanwhile, withdrew its app from the EU and UK – the first material exit driven by the AI Act compliance window for high-risk medical reference tools. Holland & Knight flagged that several US-headquartered AI companies face the same Q3 obligations on quality-management systems, post-market monitoring and incident reporting.

And here's everything else worth reading this week:

AI & tech:

• The UK government's £500m Sovereign AI Unit co-invests in Ineffable Intelligence's £830m seed round: valuing the new lab founded by AlphaGo creator David Silver at £3.8bn before it has shipped a model.

• Cohere and Aleph Alpha announce a merger to build a European frontier-model competitor: the consolidation gives Cohere European data-residency reach and Aleph Alpha access to North American enterprise channels, narrowing the trans-Atlantic gap on regulated-sector deployment.

• Anthropic launches a test marketplace for agent-on-agent commerce: the experiment puts payment rails behind autonomous agent transactions, surfacing an enterprise-AI primitive that no existing procurement framework covers.

• Microsoft integrates GPT-5.5 into Foundry as the core agentic enterprise stack: the integration positions Foundry as the governance layer Microsoft retains after losing OpenAI cloud exclusivity.

Biopharma:

• Eli Lilly and Profluent sign a $2.25bn AI gene-editing partnership targeting recombinase therapies: the deal underwrites Profluent's protein-design models with Lilly's clinical pipeline, building a data position that compounds as candidates progress through trials.

• Johnson & Johnson reports AI is halving the time required to generate drug-development leads: the disclosure is an internal R&D metric rather than a peer-reviewed result, but it sets a benchmark that competing R&D heads have to address.

• The FDA pilots real-time clinical drug trials run through cloud and AI infrastructure: A regulatory pathway for trial data to flow continuously rather than in scheduled submission windows is now clearly signalled.

• Truveta launches real-time insights drawn from real-world health data: the product packages the company's aggregation to which pharma commercial and pharmacovigilance teams can subscribe.

Medtech:

• Medtech giants are paying premium prices for AI acquisitions to defend platform position: the analysis tracks Danaher, Siemens Healthineers and Roche's diagnostics-AI deal flow, identifying a defensive consolidation playbook emerging.

• Thirty-eight critical security vulnerabilities in healthcare software found, running across 100,000 facilities: the disclosure widens the AI-security exposure for hospital IT teams already managing legacy clinical systems.

• Medtronic announces CE Mark for the Stealth AXiS surgical system: the AI-assisted spine-surgery platform clears the EU regulatory bar before its full US launch, giving European hospitals a deployment-ready system ahead of US payer-coverage decisions.

• Implicity's vendor-agnostic AI algorithm cuts false alerts on cardiac devices below the manufacturer-AI baseline: the study positions third-party AI as an additive layer over device-OEM filters, opening a hospital-procurement category that competes directly with offerings from major device makers.

• The European MedTech and HealthTech weekly tracks the regulatory and clinical-deployment news that did not surface in US trade press: the briefing covers MHRA, EMA and national-competent-authority moves that affect device classification under the EU MDR.

Advanced manufacturing:

• Eclipse closes $1.3bn across two funds dedicated to reshoring physical-industries manufacturing: the venture firm's university-endowment-backed fund directs capital at AI-equipped domestic manufacturing capacity.

• SAP's Hannover Messe demos AI use cases for manufacturing and supply-chain orchestration: the showcase moves gives SAP's enterprise-AI proposition named industrial deployments stitching planning, execution and quality into a single agentic stack.

• Wipro's partnership with Kongsberg Digital pushes Indian-services digital-twin capacity into Norwegian energy and maritime: Wipro shares rose on the announcement, reflecting investor reading that digital-twin services are repricing as deployment accelerates in heavy industry.

• Enphase develops an IQ Solid-State Transformer engineered for AI data centres: solid-state transformers compress switching gear into the same footprint as conventional units.

• Intel Foundry gains momentum as Apple eyes 18A-P and Google explores advanced packaging: the Foundry pipeline now includes Apple, Microsoft and Google as confirmed customers, repositioning Intel as the alternative US-domiciled leading-edge fab to TSMC for hyperscaler silicon.

Insurance:

• Insurity's 2026 P&C consumer survey shows AI support among policyholders has nearly doubled year-on-year: the inflection in consumer comfort removes one of the standard adoption blockers carriers cite, shifting the bottleneck back to operational and regulatory readiness.

• IMA Financial Group places a $4bn risk programme for an AI data-centre operator: the largest single AI-data-centre placement disclosed to date, structured across property, casualty and specialty layers and pricing power-density and cooling-failure exposure separately.

• Duck Creek launches an insurance-native agentic AI platform with underwriting and claims tools shipped at the same time: the platform commits a core-systems vendor to agentic AI as a product category.

• MoneySuperMarket adds home and pet insurance products to its ChatGPT app: the UK aggregator's expansion accelerates the channel-conflict question that boards have been deferring – including who owns the advice record and where the complaint path runs.

• Insurance Innovation Reporter says insurers are evaluating neuro-symbolic AI as the architecture to meet explainability requirements: do hybrid neural-and-rule systems offer a regulatory-defensible audit trail that pure neural models cannot?

• 
  

But what set podcast tongues a-wagging?

The new competitive positioning in the AI universe also cemented this week. Nathaniel Whittemore's AI Lab Power Rankings episode ran nine weighted categories and scored each lab. The operational read is that exposure to a single frontier provider carries different consequences than it did at the start of the year. Which is why a multi-lab procurement strategy is becoming the default.

First-in-class clearance is moving east.

The interview with researcher Cremieux on the a16z Show revealed China runs more clinical trials per year than the US, approves more first-in-class novel-target drugs domestically before they reach Western regulators, and has six times the number of never-before-attempted gene therapies entering trials. Which suggests first-in-class regulatory clearance is shifting eastward.

Frontier biology is cleared tissue by tissue, not platform by platform.

Peter Diamandis's interview with David Sinclair on Moonshots covered the run-up to Life Biosciences' first human epigenetic-reprogramming trial – a gene cocktail injected into the eye of a glaucoma patient to attempt to reverse blindness, with liver, kidney, ALS and macular degeneration sequenced as subsequent indications. Sinclair confirmed that the FDA is regulating the platform tissue by tissue at the agency's insistence, not as a single class clearance.

Reliability is the filter agents have to clear.

Patrick Collison's interview with the Cursor team on the a16z Show returned the conversation to how agentic deployments survive the first regulated-industry audit. Collison's reference point was Stripe's 99.999% availability across a fleet of 50 systems.

Thank you for reading this week's report. Come back next week for all the AI news you need to know in your sector.