TL;DR
The race for frontier AI infrastructure has escalated to an unprecedented scale, with Anthropic entering a massive $15 billion compute deal with SpaceX and xAI while Microsoft and Meta pivot heavily toward proprietary, closed-source architectures to protect their enterprise dominance. Simultaneously, biosecurity concerns are driving an unexpected consensus among top tech executives, who are calling for federal regulations to screen synthetic DNA as AI systems acquire advanced biological reasoning capabilities.
The Multi-Billion-Dollar Compute Alliances of Direct Rivals
The insatiable demand for frontier infrastructure is forcing direct product competitors into massive, multi-billion-dollar hardware alliances.
"...the company’s first two Colossus II clusters were built at USD 2.7 million per megawatt. That’s roughly a fourfold improvement on the industry benchmark. Paired with the USD 1.25 billion-a-month compute contract with Anthropic (disclosed elsewhere in the filing), the economics imply SpaceX recoups its AI infrastructure capex in under a month." — [A Deleted Disclosure in SpaceX’s IPO Filing Shows Its AI Infrastructure as a Cash Cow](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/anthropic-space-x-colossus-compute-deal)
As Anthropic's revenue grows, it is deploying a multi-cloud strategy that bypasses traditional rivalry boundaries to secure raw capacity, including a $15 billion deal with SpaceX and xAI for 300 megawatts of compute at Colossus 1 [Anthropic to pay xAI $1.25 billion a month in massive compute deal](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/anthropic-space-x-colossus-compute-deal). This deal illustrates that compute has become a decoupled commodity where even direct chatbot competitors must trade with one another to survive.
What to watch: Whether Anthropic's active negotiations to adopt Microsoft's custom Maia AI processors will establish a new hardware standard outside of Nvidia's dominance [Microsoft in talks to supply Anthropic with custom AI chips after $5 billion investment](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/microsoft-unveils-mai-models-build-2026).
Big Tech's Shift to Proprietary, Closed-Source Walled Gardens
Tech giants are aggressively constructing proprietary, closed-source walled gardens to claw back enterprise control and build direct commercial revenue streams.
"Bugs and infrastructure issues have pushed back the developer API for Meta's first closed-source AI model by months, with no firm launch date confirmed." — [Meta delays Muse Spark AI model API release for developers](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/meta-muse-spark-api-delays)
This pivot away from open-weight distributions—exemplified by Meta's struggling Muse Spark and Microsoft's self-sufficient MAI-Code and MAI-Thinking architectures—shows that the industry's biggest players are prioritizing monetization and direct enterprise dominance, even if it means navigating severe deployment bottlenecks [Microsoft AI CEO Says Anthropic Is Their Biggest Threat, Here's Why](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/microsoft-unveils-mai-models-build-2026). Microsoft's launch of seven homegrown MAI architectures at Build 2026 highlights a deliberate strategy to reduce its reliance on OpenAI while actively pushing its custom silicon to deliver a 30% improvement in tokens per dollar [Microsoft in talks to supply Anthropic with custom AI chips after $5 billion investment](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/microsoft-unveils-mai-models-build-2026).
What to watch: Whether Meta can successfully resolve its infrastructure bugs to launch the Muse Spark developer API in June 2026 [Meta delays Muse Spark AI model API release for developers](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/meta-muse-spark-api-delays).
Biosecurity Concerns Drive a Unified Regulatory Push
The rapid erosion of technical barriers in biological reasoning has united fierce AI and biotech rivals behind a joint push for federal biosecurity mandates.
"AI systems now outperform PhD-level virologists on questions about highly technical laboratory procedures in their own domains of expertise. The evidence about what this means for present-day biosecurity threats is genuinely mixed, but the trend is hard to dispute." — [Open Letter: In Support of Mandatory Nucleic Acid Synthesis Screening and Recordkeeping](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/ai-ceos-dna-screening-letter-congress)
On June 3, 2026, an unprecedented coalition of tech executives and gene synthesis firms issued a public open letter to US lawmakers, signaling that voluntary screening must be replaced by legally binding federal regulations [Open Letter: In Support of Mandatory Nucleic Acid Synthesis Screening and Recordkeeping](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/ai-ceos-dna-screening-letter-congress). The industry is actively asking for mandatory screening and recordkeeping, recognizing that the sheer speed of biological reasoning improvements represents a severe national security risk.
What to watch: How Congress responds to this unified industry appeal for mandatory customer screening and order tracing of synthetic genetic materials.
What surprised us
- Anthropic is bankrolling its direct chatbot rival: To secure critical hardware, Anthropic is paying SpaceX and xAI $1.25 billion per month to lease compute at Colossus 1 [Anthropic to pay xAI $1.25 billion a month in massive compute deal](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/anthropic-space-x-colossus-compute-deal).
- SpaceX's AI infrastructure is an instant cash cow: Because SpaceX built its first two Colossus II clusters at an incredibly low capital cost, the massive lease deal with Anthropic allows SpaceX to recoup its entire infrastructure capex in less than a single month [A Deleted Disclosure in SpaceX’s IPO Filing Shows Its AI Infrastructure as a Cash Cow](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/anthropic-space-x-colossus-compute-deal).
- Microsoft is targeting Anthropic over OpenAI: Despite its deep history with OpenAI, Microsoft AI's CEO Mustafa Suleyman explicitly declared Anthropic as their biggest competitive threat due to the enterprise threat of its "Cowork" coding tool [Microsoft AI CEO Says Anthropic Is Their Biggest Threat, Here's Why](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/microsoft-unveils-mai-models-build-2026).
- Meta's closed-source pivot has stalled out of the gate: While Meta spent billions to strengthen its AI capabilities, persistent software bugs have repeatedly delayed the developer API for its first closed-source system, Muse Spark, since its April launch [Meta delays Muse Spark AI model API release for developers](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70/notes/meta-muse-spark-api-delays).
Open threads worth a vote
[SpaceX and xAI Public Debut / IPO](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70#threads): Monitor the upcoming June 12, 2026 SpaceX IPO filing for further official disclosures regarding their lucrative AI infrastructure business.[Meta Muse Spark Developer API Release](/topics/019e92c9-99b4-7b6c-bb81-1e0494672f70#threads): Track whether Meta successfully launches its closed-source developer API in June 2026 or faces further infrastructure delays.