Moonshot AI Races Out Models, Deals; Eyes $2B ARR by Year-End: what are the key facts?

On September 11, Moonshot AI rolled out K2.8 Preview to all users, with capabilities approaching K3 but at roughly half the price.

Moonshot AI Races Out Models, Deals; Eyes $2B ARR by Year-End: what are the key facts?

On September 21, Kimi Code Desktop officially launched, with macOS and Windows versions released simultaneously.

Moonshot AI Races Out Models, Deals; Eyes $2B ARR by Year-End: what are the key facts?

On September 10, Moonshot AI kicked off its Moon Landing Program, partnering with five system integrators under an FDE model.

Moonshot AI Races Out Models, Deals; Eyes $2B ARR by Year-End: what are the key facts?

On September 18, Kimi K3 went live on Amazon Bedrock, putting its open-source model revenue-sharing agreement into effect.

Moonshot AI Races Out Models, Deals; Eyes $2B ARR by Year-End: what are the key facts?

Moonshot AI's ARR climbed from $100 million in March to $1 billion in August, with a year-end target of $2 billion.

Physix Frontier · News Briefing Card (enbrief · Sep 23, 2026)

Moonshot AI Races Out Models, Deals; Eyes $2B ARR by Year-End

KEY FACTS

On September 11, Moonshot AI rolled out K2.8 Preview to all users, with capabilities approaching K3 but at roughly half the price.
On September 21, Kimi Code Desktop officially launched, with macOS and Windows versions released simultaneously.
On September 10, Moonshot AI kicked off its Moon Landing Program, partnering with five system integrators under an FDE model.
On September 18, Kimi K3 went live on Amazon Bedrock, putting its open-source model revenue-sharing agreement into effect.
Moonshot AI's ARR climbed from $100 million in March to $1 billion in August, with a year-end target of $2 billion.

KEY DATA

roughly half of K3K2.8 Preview price

$3 per million tokensK3 API input price

$15 per million tokensK3 API output price

$2 billionYear-end ARR target

PHYSIX OBSERVATION

Moonshot AI is shifting from a single strong model to a dual focus on cost-performance and commercialization, but its revenue remains in the second tier. Its dense cadence of launches and revenue-sharing partnerships reflects that model companies must prove both the technical ceiling and the ability to make money, or they will struggle to justify high valuations. Whether open-source rent collection and FDE can work will determine whether it can truly close its gaps.

Source: enbrief original report ↗