Run the Numbers: CoreWeave IPO – A Future Giant or Just WeWork with GPUs?

Host: CJ Gustafson
Date: March 8, 2025
Episode Overview:
CJ Gustafson delivers an energetic, CFO-level breakdown of CoreWeave’s freshly filed S-1 as the AI hyperscaler positions itself for an IPO. He tears into revenue metrics, profitability, customer concentration, contract structures, financial engineering, and competitive threats, weighing whether CoreWeave is poised to become a foundational public AI infrastructure player or risks resembling a cautionary tale like WeWork.

Main Theme

Is CoreWeave the Next Great AI Infrastructure Company or a High-Burn Outlier with Hidden Risks?
CJ deciphers CoreWeave’s business model, explosive growth, financials, and the pitfalls and possibilities surrounding its IPO. He explains what ambitious tech operators and investors need to understand about revenue, market fit, contracts, funding structures, and the broader implications for the AI sector.

Key Discussion Points & Insights

1. What is CoreWeave and Why All the Hype?

• AI-Focused Hyperscaler: CoreWeave is a cloud provider dedicated specifically to AI workloads, renting high-performance Nvidia GPUs to tech giants and AI startups.
• Metaphor: “If AWS is a giant all you can eat buffet for cloud computing, CoreWeave is a high end sushi bar that only serves AI workloads.” (01:40)
• Revenue Model: Charges customers per GPU hour; storage sold separately.
• Customer Base: Major clients include Meta, Microsoft, and other top AI players.

2. Growth Metrics, Profitability & Revenue Visibility

• Wild Revenue Growth:
  • 2022: $16M
  • 2023: $229M — 1,346% YoY growth
  • 2024: $1.9B — 737% YoY growth
  • “They said hold my beer. We're going to grow at 737% year over year. That is not a typo.” (04:10)
• Profitability:
  • Net losses are massive but shrinking as a % of revenue:
    • 2022: –$31M (194% of revenue)
    • 2023: –$594M (259% of revenue)
    • 2024: –$863M (45% of revenue)
  • “Losses are still really big, but declining as a percentage of revenue.” (05:10)
• Contracted Revenue Dominates:
  • 96% of revenue comes from long-term contracts (2–5 years)
  • 15.1B in Remaining Performance Obligations (RPO), up 53% YoY
  • Weighted average contract duration: 4 years
  • 15-25% of contract value prepaid up front

3. Business Model Deep Dive

• “Chip Arbitrage” Middleman:
  CoreWeave buys cutting-edge GPUs from Nvidia, rents them out at a markup.
• Vulnerability: Entire business rests on continued access to top Nvidia chips—if Nvidia cuts them off or sells direct, CoreWeave is at risk.
  “If Nvidia decides to prioritize other customers or sell directly to AI companies, CoreWeave is in trouble.” (10:45)
• Take or Pay Contracts:
  Customers lock in multi-year, fixed-price deals—paying regardless of usage.
  “It's more like a gym membership for AI companies that lasts more than one year and requires really, really big prepayments.” (12:10)
• Customer Concentration:
  Microsoft makes up 62% of CoreWeave’s revenue—their contract is like a “call option on GPU demand.” (23:20)
  • “For Microsoft, not to bury the lead. This is a call option on GPU demand, guys. Cheap to hold, but with huge upside if AI workloads keep exploding.” (25:05)

4. Competition & Market Dynamics

• Cloud Giants Loom:
  AWS, Azure, and Google could ramp up and push CoreWeave out, but are currently not as AI-focused.
• Verticalization Moves:
  • Acquisition of Weights & Biases for $1.7B to boost platform stickiness and strengthen AI developer ecosystem.
• AI Funding Boom Risks:
  “If AI funding cools off, will CoreWeave's backlog hold up? Or is there a risk that some of its long term contracts don't actually get paid and companies go belly up?” (13:05)

5. Ownership, IPO Structure, and Financial Engineering

• Founders’ Control:
  Three founders (Michael Intrator, Brian Venturo, Brandon McBee) control 83% voting power via dual-class shares.
  • “If [CEO Michael Intrator] leaves, their special class B shares convert to common stock, reducing their grips. Keep an eye on if he sticks around.” (37:10)
• Heavy Institutional Insiders:
  • Investors: Magnetar, Blackstone, COATU, Nvidia (1%), Altimeter, others.
  • Magnetar has crafted convertible notes and “penny warrants,” guaranteeing sweet deals at IPO.
  • “Magnetar structured a sweetheart deal, and they're going to get shares at a steep discount. Nvidia's small stake… strategic. And private equity players—like Blackstone, COATU—may dump some shares after lockup.” (42:40)
• Off-balance Sheet Liabilities — WeWork Vibes:
  • $15B+ long-term lease commitments for data centers and power don’t show as traditional debt but behave as such.
  • “Whenever liabilities are off the books, you're like, what's going on here?” (29:00)
• Sales Efficiency:
  • Sales & marketing spend: Only $18M in 2024 (<1% of revenue)
  • “They grew revenue 737% and they only spent 18 million on sales and marketing. I've never seen something like this. That's crazy.” (35:30)
• Interest Expense vs S&M:
  • Interest expense: $332M—“18x more on interest payments than on sales and marketing.” (36:10)

6. Valuation Debate & Future Scenarios

• Key Valuation Metrics:
  • Last private valuation: $23B (Nov 2024).
  • IPO target: $35B.
  • Revenue multiples:
    • AI hardware (Nvidia): 25–40x forward revenue
    • High-growth SaaS: 15–20x
    • Cloud hyperscalers: 5–9x
  • “If they grow revenue next year, even just 100%... to 4 billion, that would still be less than 10x a forward revenue multiple.” (46:55)
• Market Positioning:
  • “If investors see CoreWeave as an AI infra pure play, it could command a premium.” (48:20)
  • “If the market starts worrying about cash burn, that's when it could get dicey.” (48:40)
• Payback Period:
  • GPUs pay off in 2.5 years; contract lengths currently 4 years.
• IPO Structures Favor Insiders:
  • “This IPO is structurally built for insiders… structured deals with key investors… insiders are shielded while public investors take on more risk.” (54:30)

Notable Quotes & Memorable Moments

• On the entire CoreWeave thesis:
  “They’re essentially playing the middleman in the AI compute economy. But as more companies try to secure their own GPUs... does this model have long term staying power?” (10:00)

• On risk of WeWork-style collapse:
  “Whenever liabilities are off the books, you're like, what's going on here? So if AI compute demand weakens, CoreWeave could be locked into expensive underutilized capacity.” (29:00)

• On Microsoft relationship:
  “They outsource AI workloads to CoreWeave… It's a call option on GPU demand. Cheap to hold, but with huge upside if AI workloads keep exploding.” (25:05)

• Sales/Marketing spend shocker:
  “They grew revenue 737% and they only spent 18 million on sales and marketing. I've never seen something like this.” (35:30)

• On Magnetar's financial engineering:
  “Magnetar’s 230 million cloud deposit looks more like financial engineering than anything. If it walks like a duck, talks like a duck, whatever the saying is…” (58:25)
  “I want a penny warrant allowing them to buy even more shares for $0.01 each. This gives them a better deal than retail investors who are buying at full price.” (1:00:30)

• Summary of strategic tension:
  “The company is growing like really fast. But its financial structure does raise the question of whether this IPO is a long term public play or just a bridge to an eventual acquisition.” (1:02:15)

Timestamps for Key Segments

• 00:00–02:30 – What is CoreWeave and "AWS buffet vs high-end sushi for AI"
• 04:00–05:15 – Insane revenue growth & the meaning behind the numbers
• 07:30–12:00 – Model dynamics: Contract backlog, RPO, prepayments
• 13:30–16:20 – AI chip land grab, chip arbitrage, and supply risks
• 20:35–26:15 – Take-or-pay contracts, Microsoft as mega-customer, concentration risk
• 29:00–31:15 – WeWork-style off-balance sheet leasing risk
• 35:30–37:30 – Sales efficiency, expense structure, interest expense spike
• 41:30–44:00 – Founders and investor ownership, dual-class share control
• 46:30–49:00 – Valuation math, comps, and market expectations
• 54:30–56:00 – IPO protection for insiders, ratchets, dilution concerns
• 58:25–1:01:00 – Magnetar’s deals and financial engineering risks
• 1:02:15–1:04:00 – Bull, bear, and base case scenarios for CoreWeave

Bull, Bear, and Base Case Scenarios

Bull Case (1:02:50)

• AI compute demand keeps exploding
• CoreWeave remains a key chip supplier, margins expand, “Snowflake for AI compute” narrative attracts premium multiples (20x+), CoreWeave emerges as AI’s hyperscaler
• “If AI model sizes and compute needs keep growing, fueling GPU demand, CoreWeave’s revenue then skyrockets beyond expectations.”

Bear Case

• AI demand normalizes, cloud giants internalize compute, contract backlog weakens, heavy liabilities, margins compress
• Valuation falls to $10–15B (5–7x forward revenue)
• “If AI compute demand normalizes and the hyperscalers eat its lunch… CoreWeave would pivot or get acquired in this scenario at a discount.”

Base Case

• High growth continues, cash burn remains, endgame uncertain, valuation $20–30B (10–15x forward revenue)
• “A highly valuable but capital hungry AI infrastructure provider that either matures into a niche hyperscaler or gets rolled up into a larger cloud play.”

Final Takeaways

• CoreWeave’s S-1 reveals staggering growth paired with high risk from customer concentration, off-balance-sheet liabilities, and insider-friendly IPO structuring.
• CJ frames the company’s future as closely linked to the AI supercycle: If demand holds, CoreWeave could be foundational; if not, it’s exposed.
• The IPO offers a playbook on insider-friendly structuring (ratchets, early liquidity, convertible deals) and the importance of understanding financial engineering before going public.

CJ’s style: Caffeinated, irreverent, candid with lots of quips (“the waffle heartburn is coming up”), tons of practical takeaways for tech CFOs, operators, and investors.

For the written version and deeper dives: mostlymetrics.com