Coinbase CEO Brian Armstrong said demand for AI is "near infinite," but expects roughly 80% of AI workloads to run on 99% cheaper models within the next 12 to 18 months.
Why it matters
The combination of those two claims is the actual signal: the demand curve keeps rising, yet the marginal cost per inference collapses by roughly two orders of magnitude. That math only works if compute and energy supply scale to absorb the volume — and Armstrong explicitly named energy and compute as the limiting factors, not chips or algorithms.
Market impact
The framing pulls energy and infrastructure plays into the AI trade rather than leaving them as a secondary derivative. For crypto, the read-through is direct: mining operators with grid access and flexible loads become natural counterparties for hyperscaler demand spikes, and AI-linked tokenization of compute capacity gets a stronger narrative spine. Investors should watch forward power-purchase agreements and data-center build-out announcements as the leading indicator of whether the supply side can actually meet the demand Armstrong is describing.
Frequently asked questions
-
What did Brian Armstrong actually predict about AI workloads?
He expects roughly 80% of AI workloads to run on models that are 99% cheaper within the next 12 to 18 months, while framing overall AI demand as near infinite.
-
Why did Armstrong name energy and compute as the bottleneck?
If demand keeps rising while the marginal cost per inference falls by roughly two orders of magnitude, the only constraint left is the physical supply of power and compute capacity.
-
How does this connect to crypto markets?
Energy-constrained mining operators with grid access can act as flexible counterparties for hyperscaler demand, and tokenized compute capacity gets a stronger AI-adjacent thesis.
-
What is the investment implication of cheaper AI models?
Cheaper inference shifts value capture away from frontier model providers toward infrastructure owners — power, data centers, and specialized compute — and toward application-layer companies that absorb the cost decline.
-
What indicator should investors watch next?
Forward power-purchase agreements and data-center build-out announcements are the cleanest read on whether supply can scale to meet the demand Armstrong is describing.
CoinDesk
Lookonchain
TheBlock
CryptoSlate
Crypto Rank News
WatcherGuru
WuBlockchain
Crypto News