🤖AI Newsletter
May 9, 2026 · 10:33 Uhr
1GPT-5.5 catapults OpenAI back to the top of AI benchmarks
@ArtificialAnlys / TikTok @thelastweekinai GPT-5.5 (codename 'Spud') surpasses Anthropic and Google on the Artificial Analysis Intelligence Index by 3 points, ending a three-way tie. Additionally, OpenAI released GPT-Realtime-2, which ranks 1st in the Speech-Reasoning benchmark with 96.6%. OpenAI thus establishes a clear technical lead, while prediction markets still see Anthropic as the favorite for late May – a fascinating contradiction.
2SpaceX plans $55 billion chip factory 'Terafab' for AI self-sufficiency
New York Times SpaceX is investing 55 billion dollars in its own AI chip factory called Terafab, intended to supply Tesla and SpaceX with proprietary semiconductors – independent of Nvidia or TSMC. The project underscores the trend toward vertical integration in AI hardware and could significantly destabilize the chip market. Tax incentives are being requested, providing fodder for political debate.
3Meta monitors employees comprehensively for AI training data
New York Times According to the NYT, Meta tracks all computer inputs, mouse movements, and screen contents of its employees to train AI models with real work data. This practice is increasingly dissatisfying employees and raises serious questions about data privacy and worker rights. The case exemplifies the growing conflict between aggressive AI data strategies and corporate culture.
4Pentagon contracts divide AI industry: Google yes, Anthropic no
TikTok @mattrwolfe / @thepromptculture The US military awarded classified AI contracts to OpenAI, Google, and xAI – Anthropic declined and faces pressure as its safety positioning is being commercially penalized. Subsequently, Google DeepMind employees in the UK voted for unionization, explicitly citing the Pentagon deal as a trigger. The split shows that AI safety policy is increasingly becoming a real competitive disadvantage.
5Anthropic automates Wall Street analysts with 10 finance agents
@JoshKale / winbuzzer.com Anthropic released ten specialized Claude agents for the financial sector, including Pitch Builder, Meeting Preparer, and Deal Analyzer – directly entering the core domain of investment banking analysts. Simultaneously, Anthropic and OpenAI jointly launch enterprise AI service ventures, dramatically intensifying B2B competition. A developer commented on the Claude Managed Agents launch with: 'There goes a whole YC batch.'
Situation Report
In May 2026, the AI industry is in a phase of accelerated power concentration: OpenAI reconquers the technical top position with GPT-5.5, while SpaceX aims to break Nvidia hardware dependency with a 55-billion-dollar chip factory – both signal a new round in the AI infrastructure arms race. Simultaneously, the conflict between commercial interests and ethical principles is escalating: Meta's comprehensive employee monitoring for AI training and the Pentagon contract awards to Google and xAI show that safety boundaries are being systematically shifted under economic pressure. Anthropic's refusal to cooperate militarily costs the company market share but demonstrates a strategic differentiation logic that could bring regulatory advantages in the long term. For companies, the situation means: whoever does not control AI infrastructure themselves and does not actively implement agent automation is already structurally losing competitiveness against early adopters today.
Tokens: 2,112(1,296 in · 816 out)
