The numbers are staggering. $2.575 billion in revenue. 112% growth year-over-year. Five Wall Street analysts raising price targets, with Wells Fargo punching $160. CoreWeave’s Q2 earnings are being celebrated as a victory lap for AI infrastructure. But behind the headlines, a deeper question gnaws at the edges of this narrative: what does CoreWeave’s success really tell us about the state of AI compute—and the illusion of digital sovereignty we’re buying into?
I’ve spent years translating cryptographic concepts for institutional executives. I’ve seen how quickly they embrace the language of ‘sovereignty’ and ‘transparency’ when it suits their portfolios. But CoreWeave isn’t a decentralized protocol. It’s a centralized cloud provider, heavily dependent on NVIDIA’s GPU supply chain and a handful of hyperscale customers. Its earnings report, while financially impressive, is a black box of opacity. No code to audit. No governance to scrutinize. No token to align incentives. Just a stock price and a promise of more megawatts.
This is the context we must hold close: the AI compute market is booming, but the infrastructure remains centralized, fragile, and opaque. For the Web3 faithful, CoreWeave is a stark reminder of what we’re fighting against—and a data point that the demand for compute is real, even if the delivery mechanism is flawed.
Let’s dive into the core of the report. The headline revenue growth is undeniable. Analysts point to ‘net new active megawatt growth’ and ‘robust Q3 bookings’ as signs of sustained demand. Piper Sandler raised its target to $153, Baird to $130. The market is pricing in a future where AI workloads continue to explode. But here’s the rub: two major institutions—JPMorgan and Mizuho—remain neutral, with targets of $120 and $115 respectively. Their concern? Short-term revenue uncertainty. They see the growth, but they question its durability. Based on my experience auditing smart contracts for EthicChain in 2017, I learned that transparency is the primary mechanism for trust. CoreWeave offers none of that. We don’t know its customer concentration, its gross margin breakdown, or the terms of its supplier contracts. We are asked to trust, not verify.
Speed kills. Precision saves. The market is moving fast, but it’s not moving precisely. The divergence in analyst targets—from $115 to $160—signals a fundamental disagreement about the sustainability of this growth. Some see a virtuous cycle of AI adoption; others see a capex bubble inflating on borrowed time. The truth likely lies in the middle, but the absence of verifiable data makes it impossible to know. Audit the algorithm, not just the code. CoreWeave’s algorithm is its supply chain, its customer relationships, its capital allocation. None of that is audited by the public.
Here’s the contrarian angle: CoreWeave’s success is actually a cautionary tale for the AI x Crypto narrative. The very strengths it boasts—scale, centralized control, reliance on a single chip supplier (NVIDIA)—are the vulnerabilities that decentralized compute networks (DePINs) are designed to solve. After the Terra collapse, I retreated to a Bali cabin to process the trauma. That solitude taught me to look beyond yield and focus on resilience. Today, I see the same hubris in centralized AI cloud providers. They promise infinite scalability, but they are captive to hardware cycles, geopolitical risks, and corporate decision-making. The next crisis won’t be a bank run; it will be a GPU shortage, a power outage, or a geopolitical export ban that cuts off a nation’s compute supply.
Trust no one, verify the solitude. CoreWeave’s report validates that AI compute demand is real, but it also validates the need for decentralized alternatives. Projects like Render Network, Akash, and others offer a different path: compute that is verifiable, permissionless, and resilient. They may lack the scale of CoreWeave today, but they offer something Wall Street can’t price: sovereignty. The institutional translation work I did in 2024 taught me that compliance is not censorship—it’s transparent accountability. CoreWeave’s model is neither transparent nor accountable. It is a black box that happens to be profitable.
Looking forward, the market is pricing the wrong thing. It’s pricing GPU access, not compute sovereignty. The next cycle will not be about who has the most megawatts; it will be about who can provide verifiable, decentralized compute that is immune to single points of failure. The Q3 bookings may be strong, but the real signal is not in the numbers—it’s in the architecture. Speed kills. Precision saves. The race is on, but the finish line is not Wall Street’s approval. It’s the preservation of human agency in an algorithmic age.
Are we building infrastructure for liberation, or just a faster cage?