AI news5
Anthropic-SpaceX deal extends to 220,000 Nvidia GPUs and orbital data centre exploration
Follow-up reporting on 7 May 2026 confirmed the Anthropic-SpaceX Colossus 1 deal includes access to more than 220,000 Nvidia GPUs across the 300MW Memphis facility, with capacity ramping in within a month. The agreement also includes a research-stage commitment to study multi-gigawatt orbital AI compute. Anthropic immediately raised Claude Pro, Max, and Opus API usage limits.
spacenews.com ↗Why it matters
A frontier lab pre-committing to terrestrial GPU and orbital compute at the same announcement reframes "compute scarcity" as a strategic positioning question rather than a near-term technical bottleneck.
DeepMind launches AlphaEvolve, a Gemini-powered coding agent for science and infrastructure
Google DeepMind announced AlphaEvolve on 7 May 2026, a Gemini-powered coding agent designed to scale impact across science, infrastructure, and research. The system is positioned to write and improve algorithmic code through evolutionary search, with case studies in scientific simulations, hardware optimisation, and research workflows. DeepMind framed it as the first agent shipping under their applied-research delivery model.
deepmind.google ↗Why it matters
A frontier-lab coding agent tied to evolutionary search and applied-science deployments narrows the gap between research demos and shipped tools, the bar nonprofits should hold their own AI tooling to.
Frontier Model Forum names Chris Meserole as first Executive Director
The Frontier Model Forum, the cross-lab governance body backed by OpenAI, Anthropic, Google, and Microsoft, announced Chris Meserole as its first Executive Director on 7 May 2026. The appointment formalises the Forum's governance structure as the labs face increasing regulatory and pre-deployment-evaluation expectations. Meserole comes from Brookings, where he led AI policy work.
lesswrong.com ↗Why it matters
A formalised governance body with a named ED gives regulators and counterparties a single industry interlocutor, which will gradually shape how AI safety norms propagate downstream to vendors and nonprofit users.
US Department of Defense approves 8 AI companies for classified-network use
The DoD announced on 5 May 2026 that eight AI companies are now approved to operate on classified Pentagon networks, expanding an approved-supplier list that previously held only a handful of vendors. The authorisation covers operational AI deployments, not just research access. Combined with the CAISI pre-deployment evaluation programme, US frontier-AI vendors are now explicitly working both consumer-protection and national-security tracks.
advancedmanufacturing.org ↗Why it matters
Defense and national-security AI procurement is normalising at scale, which sets the institutional bar for AI governance that nonprofit AI vendors and partners will eventually face.
OpenAI board reportedly discussed merging with Anthropic during "the Blip"
The Verge reported on 7 May 2026 that OpenAI's board discussed a potential merger with Anthropic during the November 2023 leadership crisis known as "the Blip", and considered Dario Amodei as a possible OpenAI CEO. The disclosure surfaced through Musk v. Altman trial exhibits. The two companies have since followed parallel paths into Wall Street joint ventures, frontier model release races, and CAISI pre-deployment programs.
theverge.com ↗Why it matters
A near-merger between the two largest frontier labs in late 2023 reframes the current independent-but-converging strategic posture, and reminds counterparties that AI vendor consolidation remains a live possibility.
AI in the nonprofit sector4
Rockefeller Foundation, Universidad Icesi, and City of Cali deploy Dengue.AI for outbreak prediction
The Rockefeller Foundation, Universidad Icesi, and the City of Cali signed an MOU to deploy Dengue.AI, an AI system that predicts dengue outbreaks up to three weeks in advance and prescribes targeted interventions. The deployment covers approximately 2.2 million people in Cali, Colombia. Announced 5 May 2026.
rockefellerfoundation.org ↗Why it matters
A frontline public-health AI deployment with a major foundation, an Andean university, and a municipal government as named co-signers is the operational pattern other foundation-funded AI-for-good initiatives will increasingly follow.
NextLadder Ventures launches $1B+ NavTech fund for life-decision support tools
NextLadder Ventures unveiled its inaugural plan on 6 May 2026, backed by over $1 billion in capital, to invest in "Navigation Technology" or NavTech: tech-enabled support tools that help Americans navigate major life decisions in education, employment, and economic mobility. The fund announced its co-founder leadership team and named investment focus areas spanning workforce, education, and benefits-access AI.
aijourn.com ↗Why it matters
A $1B-scale fund explicitly framing life-decision-support tooling as a category, rather than a feature, validates the AIH-orbit "AI as practical scaffolding for under-resourced individuals" thesis at venture-grade scale.
American Express launches AI Upskilling for Small Business + Generation + Scholarship America partnership
American Express announced two initiatives on 6 May 2026 in partnership with Generation and Scholarship America to support nonprofit and small business AI adoption. AI Upskilling for Small Business (with Generation) is a global, practitioner-focused program; the Scholarship America partnership funds AI-skill scholarships for under-resourced learners. Specific dollar amounts and eligibility windows are on the AmEx newsroom page.
businesswire.com ↗Why it matters
Corporate-philanthropy-plus-nonprofit-implementation-partner is a model that already works in Australia; the AmEx programme is a useful template for any AU corporate sponsor conversation.
ImpactAI Foundry launches in Bangalore to build AI capacity in India's nonprofit sector
ImpactAI Foundry launched on 7 May 2026 as a hands-on AI capacity-building program for India's nonprofit sector. The June 2026 pilot cohort will partner six to eight Bangalore-based nonprofits with volunteer tech mentors from the local ecosystem to identify and ship AI deployments inside the orgs. The model emphasises mentor-pairing over training-only delivery.
cxotoday.com ↗Why it matters
A direct India-based peer to AIH's education-first AI-for-nonprofits model is now in market, which is both a comparison point for AIH's positioning and a candidate for a peer-to-peer learning conversation.