Martin Casado says AI has erased startups' distribution barrier

In the a16z interview, Casado said, "AI just solves the distribution problem. It solves the demand problem." Getting a new service noticed and adopted used to be the hardest part of running a startup, but now demand for tokens and GPUs is so large and so steady that all a company has to decide is how much money to spend — and that spending translates directly into top-of-funnel growth. As a result, Casado explained, the size of the funding rounds startups can raise has grown to the point where they can hold their own against Microsoft and Meta in a straight capital contest.

Why he named Cursor, Anthropic, and OpenAI

The three companies Casado named as examples of explosive growth are Cursor, Anthropic, and OpenAI. Cursor started out as an AI coding editor that let users pick between Anthropic's Claude, OpenAI's GPT, and Google's Gemini right inside the app, while also training its own model, Composer, on the side. In August 2026, SpaceX acquired Cursor, bringing it under the same roof as xAI. Microsoft is on a different track: its consumer AI unit, Microsoft AI, licenses OpenAI's models for Copilot, Bing, and Edge while separately training its own model, MAI. The gap Casado is pointing to isn't a difference in org structure — it's a shift in fundraising power, where startups can now deploy capital at a scale that used to be reserved for incumbents.

The stakes grow as prices fall

As covered in OpenAI, Anthropic launch price war as Chinese AI rivals close the gap, OpenAI cut both input and output token prices for GPT-5.6 Luna by 80% this past August, and Anthropic priced Claude Opus 5 at half the cost of its flagship model, Fable 5. Cheaper tokens mean the same budget buys more inference, which dovetails with the "unlimited demand" logic Casado describes.

Model	Input price change	Output price change
GPT-5.6 Luna	$1 → $0.20 (down 80%)	$6 → $1.20 (down 80%)
Claude Opus 5	Half of Fable 5 ($5)	Half of Fable 5 ($25)

As per-token costs fall, the scale of service a startup can build on the same budget keeps growing. Flip Casado's argument around, and the point where big tech's capital advantage stops mattering turns out to be tied directly to the ongoing decline in AI compute costs.

Editor's take

Taken at face value, Casado's comments could be read as "capital no longer matters" — but it's more accurate to say the way capital gets used has changed. Incumbents used to be able to push out newcomers by leaning on distribution networks and sales organizations built up over years. Now, in a market where token prices keep falling and GPU demand shows no sign of slowing, investment dollars convert directly into compute and, from there, into a growth curve. That's why a company like Cursor could grow valuable enough for SpaceX to acquire it — as Casado puts it, the story isn't about distribution, it's about fundraising capacity.

Comparing generations of startups makes this more tangible. A few years ago, startups had to pour their entire early runway into building growth metrics, and half their resources went into just cracking an incumbent's existing customer base. Now, plenty of companies skip that whole phase simply by plugging in a model whose per-token cost has dropped sharply from where it used to be.

The practical takeaway for Korean startups is straightforward. Rather than blindly ramping up distribution and marketing budgets, recalculating your cost structure every time token prices drop — and reinvesting the savings into user acquisition — builds a faster growth curve. That logic still only holds for products that already have real users engaging with them: if there's no underlying demand, burning cash won't create the "unlimited demand" Casado is describing.

In the coming weeks, this debate will likely get tested again in the size of the next funding rounds for Cursor, Anthropic, and OpenAI. If token prices drop even further, the funding contest between startups and incumbents should narrow even more — just as Casado predicts.