While AI firms like Anthropic and OpenAI are trying to build God, Google (GOOGL) is out here building a business model.

The race for artificial general intelligence (AGI) is dominating the discussion for most AI players, and it’s yielding unwieldy results. Anthropic recently warned that its unreleased Mythos model might be too powerful for human civilization.

Gemini-parent Google, however, is taking a different approach.

Sundar Pichai’s pitch isn't bigger brains, it's a smaller bill. The firm’s new Gemini 3.5 Flash model claims to rival frontier offerings while saving companies money as they burn through billions of tokens running AI agents.

Pichai recently noted that monthly usage of its AI products has increased sevenfold to 3.2 quadrillion tokens since last year. If Google Cloud's biggest customers shifted 80% of their workloads to a mix of Flash and other frontier models, he said, they could save more than a billion dollars a year.

You don't have to win the race to AGI if you own the road everyone else is paying to drive on.

— Brooks & Cas

What: OpenAI CEO Sam Altman may have beat SpaceX’s Elon Musk in court, but the latter may have the edge on Wall Street.

Why: SpaceX filed for an IPO targeting roughly $1.5T, with a listing expected mid-next-month, while OpenAI's own filing is reportedly imminent off an $852B private mark. SpaceX generated $18.7B in 2025 revenue against OpenAI's $13.1B, both middling for the S&P 500's trillion-dollar club, which averages north of $260B. But only one has a true moat.

What Else: SpaceX accounts for more than 80% of the world's mass to orbit since 2023, a position Roth Capital's Rohit Kulkarni calls monopolistic. Starlink runs roughly 9,600 satellites, about 15x the next-largest fleet. OpenAI has no such grip. ChatGPT is still huge, but Anthropic's Claude and Google's Gemini are eating its first-mover lead, and training costs are expected to swallow its revenue for years. Meanwhile, SpaceX just started renting compute to Anthropic rather than hoarding it, and the latter on track to achieve its first-ever profitable quarter.

Watch: Starship's next test after last week's booster trouble. When the launch monopoly stumbles and barely flinches, you're watching a moat.

What: Blue Origin's New Glenn rocket burst into flames during a hotfire test at Cape Canaveral. Cause unknown, personnel safe. However, the shrapnel did hit some stocks.

Why: AST SpaceMobile (ASTS) slumped on Friday, since the blast pushes out the billion-dollar backlog it owes carriers like Verizon and AT&T. But there could be a bigger name in the blast radius. Amazon (AMZN) is attempting to replicate SpaceX’s Starlink success, and racing to launch enough satellites to keep its spectrum licenses. It’s leaning on Blue Origin, also founded by Jeff Bezos, to do so.

What Else: Rocket Lab (RKLB) and Firefly actually own working launch capability, which should be worth more when rivals can't reach orbit. Both fell sharply anyway. The selloff says less about rockets than about how stretched these names got ahead of the SpaceX IPO.

Watch: SpaceX, which handles over half of all orbital launches and just watched its competition combust. Less rivalry for Starlink, more reason to take profits up here.

What: Wall Street will pay billions for the mere promise of powering AI someday. Fervo Energy (FRVO) actually has a date on the calendar. The geothermal firm debuted this month and trades up more than 40% from its IPO price at a $12.4B market cap, roughly level with nuclear names Oklo (OKLO) and X-Energy (XE) at around $11.5B each.

Why: The difference is proximity. Fervo's first Cape Station unit in Utah is slated to deliver power by Oct. 1, two more by Jan. 1, with penalties if it misses. Most nuclear SMR startups don't expect commercial delivery until 2030, and many lack NRC design certification. Fervo holds 658 megawatts, or $7.2 billion, in binding deals with Google, Shell (SHEL), and others.

What Else: There's also a cost curve doing the heavy lifting. BloombergNEF says Fervo's per-well drilling costs fall about 29% with each doubling of wells drilled. Project InnerSpace pegs later-phase returns at 15% to 21%, beating the single digits solar and wind typically deliver and rivaling Permian shale. Always-on baseload is exactly what hyperscalers can't get from gas turbines now priced near triple 2019 levels.

Watch: Whether Fervo's wells hold steady heat across 15-year contracts. The pilot only dates to 2023, so the thesis is promising, but unproven.