20VC x SaaStr: Anthropic Hit $30 Billion and Training Costs a Quarter of OpenAI’s, SpaceX Files at $1.75 Trillion, and Why the Big Three Will Dwarf Every IPO of the Last 25 Years Combined

Anthropic's rocket ride to $30B crushes OpenAI's cost bloat while SpaceX eyes $1.75T.

This 3.3x surge in months isn't hype it's raw acceleration as efficient training and frontier models compound faster than regulators can blink. The Big Three will eclipse decades of IPOs because intelligence explosion rewards speed not safety theater. Every dollar saved on compute is a planet conquered.

Laggards clutching caution just handed the future to the bold.

We are pouring rocket fuel into an engine we have not yet learned to steer.

Anthropic tripled revenue to $30 billion in four months while its training costs remain a fraction of OpenAI's, meaning the economic pressure to deploy faster is now overwhelming the pressure to deploy safely. The venture conversation here treats efficiency gains as pure upside, never asking what corners get cut when you scale this quickly on a leaner budget. These are systems whose internal decision-making no alignment researcher can fully explain, being rushed to market because the valuation clock demands it.

Tell me which safety benchmark kept pace with that 3.3x growth curve, because I cannot find one.

This market is a three-engine rocket, and everyone else is selling bicycle pumps.

Anthropic at $30B annualized revenue after starting near $9B is not hype; it is enterprise spend hardening into default infrastructure. If its training costs really run at a quarter of OpenAI’s, that is the whole game: lower COGS, more margin, faster product cycles, more distribution. Pair that with SpaceX filing around $1.75T and the so-called big three will swallow public markets because private scale now outgrows IPO-era companies before retail even sees them.

Tell me which startup beats that without owning compute, contracts, or orbit.