Bloomberg Tech: Google to Release New AI Chips (Podcast)

Google plans to release new AI chips focused on inference, directly challenging Nvidia in the AI hardware market. The segment also noted Blue Origin’s successful booster recovery but failed satellite orbit insertion, which sent AST SpaceMobile shares lower, and Cerebras’ plan to pursue an IPO after previously withdrawing a listing attempt.

Analysis

GOOGL’s inference-chip push is strategically less about displacing Nvidia in the near term and more about compressing the economics of frontier model serving inside Google’s own ecosystem. The first-order beneficiary is Google Cloud/YouTube/search monetization, where lower inference cost can widen margins or fund more aggressive pricing; the second-order risk for NVDA is not immediate unit loss, but a slower growth algorithm as hyperscalers keep more workloads on proprietary silicon and reserve Nvidia for training and edge cases.

The market should not overread this as an all-clear for GOOGL or a structural top for NVDA. Custom silicon only matters if software, scheduling, and compiler maturity can sustain performance gains across diverse model architectures; that typically takes quarters, not weeks. If Google’s new chip is inference-only, it still reinforces a bifurcation: Nvidia remains the default for training capex, while Google, AWS, and Microsoft increasingly compete on serving efficiency, which caps Nvidia’s mix expansion and pricing power on mature workloads.

For ASTS, the negative impulse is more about execution credibility than the satellite mishap itself. In this part of the cycle, investors are paying for de-risked launch cadence and orbit insertion reliability; a single visible failure can widen the cost of capital quickly because it raises the probability of schedule slippage on a capital-intensive buildout. The next 1–2 launches are critical: if they normalize, the stock can recover sharply, but if reliability doubts persist, financing terms and partner negotiations likely become the real damage channel.

Contrarian takeaway: the Nvidia reaction may be too blunt if the street is extrapolating a multi-year share loss from a chip announcement that mainly improves Google’s internal economics. The more interesting trade is that hyperscaler capex budgets may shift from general-purpose AI spend to infrastructure efficiency, which can actually support total AI spend even as it changes who captures the margin. ASTS looks more vulnerable tactically, but the selloff may be overdone if this was an isolated launch miss rather than a pattern.

Key Decisions for Investors

• Short NVDA tactically into any strength over the next 1-2 weeks, but keep size modest; the best risk/reward is against inference-related multiple compression, not a fundamental earnings collapse.
• Long GOOGL vs short NVDA as a 3-6 month pair trade; thesis is margin capture at the hyperscaler layer, with upside if Google proves the chip improves serving economics and cloud competitiveness.
• For ASTS, wait for confirmation rather than catching the knife: buy only after the next successful launch/orbit milestone, or express a bearish view via near-dated put spreads if volatility remains elevated.
• If you want to own AI infrastructure but reduce Nvidia-specific risk, rotate into a basket of hyperscalers and networking names that benefit from lower inference costs without depending on chip OEM share.