Jan 7, 2026 · 59m · big-technology
Coreweave: AI Bubble Poster Child Or The Next Tech Giant? — With Michael Intrator and Brian Venturo
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In this episode of the Big Technology Podcast, host Alex Kantrowitz interviews CoreWeave co-founders Michael Intrator and Brian Venturo to explore how the specialized cloud provider designs custom AI data centers, manages capital debt risk, and scales compute infrastructure amidst the global AI boom.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. Alex holds 20.6% of the talking time here. How this is scored →
Alex as informed peer 5.7 Guest teaching 6.2 Guest disagreement 3.7 Alex pushing back 3.9
speaking balance: gold is Alex, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 46:09 Michael mocks short sellers and commentators on GPU depreciation
Michael forcefully dismisses the depreciation narrative by mocking commentators who couldn't spell GPU two years ago but now present themselves as experts.
Hardest push from Alex ▶ 44:39 Alex insists on finishing his counterargument regarding GPU obsolescenceWhen Michael tries to jump in early, Alex firmly holds the floor to articulate the full risk of older GPUs depreciating rapidly as models shift to inference.
Biggest teaching moment ▶ 36:19 Michael details the SPV box and waterfall structure to debunk debt rumorsMichael breaks down structured finance mechanics step by step, showing how customer revenue directly funds operating expenses and interest inside an isolated box before equity release.
Alex holds their own ▶ 25:54 Alex challenges CoreWeave on avoiding becoming a distressed asset during a downturnAlex cuts directly through the founders' optimism about market downturns by questioning their structural safeguards against becoming the distressed asset themselves.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | Alex as informed peer | Guest teaching | Guest disagreement | Alex pushing back | Why |
|---|---|---|---|---|---|---|
| Hypergrowth, Operational Friction, and Internal Culture | 4 | 3 | 1 | 2 | Alex sets the stage with current valuation and datacenter deployment stats before asking about the operational strain of the AI boom. Michael and Brian describe the emotional and logistical reality of scaling from 100 to 2,500 employees. | |
| Origins from Crypto Mining to Diversifying Customer Backlog | 5 | 6 | 3 | 3 | Alex brings up CoreWeave's crypto mining roots and claims customer concentration in Microsoft is two-thirds of their backlog. Michael immediately corrects him with recent earnings data showing no single customer accounts for more than 30%. | |
| Engineering Differentiation and Proprietary Data Center Architecture | 5 | 6 | 2 | 2 | Alex asks about the engineering required to cool high-density chips, and Brian clarifies that cooling hardware is commoditized while software automation is the real differentiator. Michael elaborates on how CoreWeave decommoditizes GPU infrastructure compared to other neoclouds. | |
| The Shifting Balance Between AI Training and Inference | 5 | 4 | 1 | 2 | Alex asks whether customer compute usage has shifted from model training toward inference workloads. Brian explains how inference compute now mirrors training infrastructure and notes their workload split has moved close to fifty-fifty. | |
| Why CoreWeave Exists: Parallel Computing vs. Cloud 1.0 | 6 | 6 | 3 | 5 | Alex challenges the founders on CoreWeave's core value proposition versus tech giants building data centers internally. Brian and Michael argue that the shift from sequential Cloud 1.0 to parallelized Cloud 2.0 requires completely different architectural specialization. | |
| Risk Management, Long-Term Contracts, and Option Value | 6 | 7 | 5 | 5 | Alex puts forward the bear argument that hyperscalers use CoreWeave simply to offload capital risk. Michael rejects this narrative as model fitting, breaking down their risk allocation into 85% fully contracted debt-backed infrastructure and 15% future option value. | |
| Debt Structuring and Cash Flow Waterfall Mechanics | 7 | 8 | 6 | 6 | Alex raises sharp analyst critiques regarding debt loads, OpenAI commitments, and rumors of borrowing to service existing interest. Michael systematically explains the bankruptcy-remote SPV box structure and cash flow waterfall mechanics to demonstrate how contracts fully amortize the debt. | |
| Hardware Lifecycles, Depreciation, and GPU Obsolescence Myths | 7 | 8 | 7 | 6 | Alex presses on hardware obsolescence, arguing newer GPUs will rapidly erode the value of older fleets. Michael and Brian push back aggressively, mocking short sellers who couldn't spell GPU two years ago and citing multi-year A100 renewals executed at 95% of original contract values. | |
| Addressing NVIDIA Ties and Circular Financing Allegations | 6 | 7 | 6 | 5 | Alex asks directly about circular financing allegations stemming from NVIDIA's equity stakes and compute leasing deals. Michael contextualizes NVIDIA's few hundred million in equity against CoreWeave's $25 billion raised and $40 billion valuation, calling the theory ridiculous. | |
| Energy Grid Bottlenecks, Construction Labor, and Efficiency Gains | 6 | 7 | 3 | 3 | Alex cites Satya Nadella's comments on power availability being the primary bottleneck for data center expansion. Brian reframes the issue, explaining that skilled construction trade labor and energization testing are the true near-term constraints rather than generation capacity. |
Statements from this episode (20)
Venturo: CoreWeave grew from 100 to 2,500 employees in three years
“Where we had maybe a hundred employees three years ago, now we have 2500 employees or so”
Intrator: No single customer accounts for over 30% of CoreWeave backlog
“There's no customer that represents more than 30% of our backlog, and so we've done an incredible job.”
Venturo: Liquid cooling differentiation is in software control, not physical plumbing
“Everybody believes that this, that there's some differentiation in the plumbing of the liquid cooled data center, right? That's not where the differentiation lies. It's all the same pipe and valves and fittings. Like everyone's using the same things there. The…”
Intrator: CoreWeave's deployed GPU fleet exceeds all other neoclouds combined
“If you took all of the other neoclouds and added their GPU fleets up, we would still be a multiple of all of them combined in terms of the number of GPUs that are up and running and delivered to clients.”
Venturo: CoreWeave owns the largest dataset on AI infrastructure failures
“We have the world's largest data set of how these things run, how they fail, and we've built all the recovery mechanisms and the software intelligence to help our customers run these things.”
Venturo: No difference between training and inference infrastructure for modern AI
“So there's really no difference between training infrastructure we deployed to build those capabilities and what our customers are ultimately using to serve them.”
Venturo: CoreWeave compute demand has shifted to roughly 50-50 training and inference
“Six months ago, I would have said it was two-thirds training and one-third inference. It's probably close to fifty-fifty now.”
Intrator: Google is building $50B in infrastructure while buying external capacity
“Google went out and just You know, released a press release where they're building fifty billion dollars worth of infrastructure while they're still buying from everyone else they can. Microsoft is building internally, and they're buying from lots of other pla…”
Intrator: CoreWeave secures 85% of infrastructure exposure with long-term contracts
“What we are doing Is we are making the majority of our investments by taking Long-term contracts from credit-worthy entities using those contracts as a way of raising money to build the infrastructure where the demand and the credit and the capital has already…”
Intrator: CoreWeave fully amortizes debt and profits within 5-year contracts
“Within the five years of the contracted period with Microsoft, we pay for the infrastructure, we pay for the OPEX, we pay for the interest, and we earn an enormous margin on the infrastructure.”
Intrator: OpenAI has 800 million monthly active users
“They have a franchise that has eight hundred million monthly users of their product, which is fully one 10th, one out of every 10 human beings on the planet logs on to OpenAI.”
Intrator: CoreWeave signed a $14 billion contract with Meta
“Meta that we signed a fourteen billion dollar contract with”
Intrator: CoreWeave borrowing spread dropped from SOFR+1350 to SOFR+400
“No, it was SOFR plus 13 50 down to SOFR plus 400, right?”
Venturo: Hyperscalers ran NVIDIA K80 fleets for 10 years before retiring them
“Last year is when we saw let's call it the hyperscalers that were around in the twenty-tenths, so Amazon, Microsoft, and Google, finally retire their NVIDIA K-eighty fleets, and the K-eighty was a GPU that was introduced in 2014. So it's, it was active in thei…”
Venturo: CoreWeave GPUs deployed in 2019 remain running with active customers
“The idea that these things burn out in two or three years, like, it's kind of bunk, right? And from a physical perspective, Right. Within three years, these things are all still under warranty. So if they break, they get replaced. Right. But from a, like, this…”
Venturo: CoreWeave re-contracted 5-year-old NVIDIA A100s at 95% of original pricing
“In addition to the H 100, which came out in 2023, right, we signed a term contract for the A 100, like within like the 95% of an original price range for on, like on term last week or two weeks ago. Like, that's crazy. Those GPUs are already five years old and…”
Intrator: NVIDIA invested $350M across two rounds while CoreWeave raised $25B
“NVIDIA has made two investments in CoreWeave. They made an investment of a hundred million dollars. And then they made an investment of and that was early, that was in the B round, I believe. .
Brian Venturo: At a two billion dollar valuation.
Michael Intrator…”
Intrator: CoreWeave is worth $40 billion
“From the perspective of, you know, the company is worth forty billion dollars.”
Venturo: Construction labor, not power, is currently limiting AI data center builds
“And right now I don't think that power itself, meaning grid connections and the generation capacity is the limiting factor. Right now it's the construction and trades. So it's human labor and supply chain that are the limiting factor is that you went from a ma…”