Thursday, 30 July 2026

The AI investment thesis is fracturing. Amazon and Microsoft are thriving on the back of sustained cloud spending for AI infrastructure, with Amazon reporting 37% AWS revenue growth and $220 billion in capex commitments for 2026, while investors show clear skepticism toward AI labs and startups without dependable revenue streams. Situational Awareness, the hedge fund founded by Leopold Aschenbrenner, exemplifies the risk: after betting heavily on AI memory chips and energy plays, it's liquidating most of its public portfolio after losing roughly half its assets in recent months, retaining only its $5 billion Anthropic stake in hopes of October's expected IPO.

Meanwhile, the market is revealing where AI's real friction points lie. OpenAI aggressively cut prices on GPT-5.6 models—Luna dropping 80 percent three weeks after launch—signaling intensified competition from Chinese competitors on cost rather than capability alone. Reddit's warning about "choppy" search referrals and a slight decline in daily actives shows AI-powered search summaries are quietly cannibalizing referral traffic, a dynamic that could destabilize content platforms relying on data licensing deals. Google DeepMind shipped Gemini Robotics 2, enabling multi-robot coordination and cross-embodiment skill transfer with 32 to 92 percent success rates depending on task complexity, suggesting AI's next frontier moves from inference to physical control.

Security is becoming the unglamorous foundation of AI's future. Google fixed 1,072 Chrome bugs in two June releases using internal AI tools—more than the prior 23 releases combined—while enterprises are struggling to secure AI-generated code lacking confidence in their own systems. Okta's $200 million acquisition of Permiso and the OpenAI-Elastic partnership on agentic safety reflect a market realization: controlling what AI systems access and do matters more than raw model performance.