Yazılar

CoreWeave to Spend Up to $23 Billion in 2024 to Fuel AI Data Center Expansion

/in /tarafından

Nvidia-backed CoreWeave plans to invest between $20 billion and $23 billion this year to build out AI infrastructure and data center capacity, the company announced Wednesday, citing surging demand from clients like Microsoft and OpenAI.

Despite reporting better-than-expected Q1 revenue in its first earnings report since going public in March, shares fell 5%, as investors reacted to the company’s aggressive capital expenditure plans.

“You pay back your infrastructure fully loaded and you have significant profit,” said CEO Mike Intrator, defending the strategy in an interview with Reuters.

📊 Financial Highlights

  • Q1 Revenue: $981.6 million (vs. $852.9M expected)

  • Q2 Revenue Guidance: $1.06B – $1.1B

  • Q2 CapEx Forecast: $3B – $3.5B

  • Full-Year Revenue Forecast: $4.9B – $5.1B (above $4.61B analyst consensus)

  • Revenue Backlog (as of March 31): $25.9 billion

🔌 AI Infrastructure Boom

CoreWeave operates high-performance data centers tailored for AI model training and inference, powering major clients like:

  • Microsoft, which uses CoreWeave as overflow compute capacity

  • OpenAI, with whom it signed an $11.2 billion five-year deal in March that also includes an equity stake for the ChatGPT maker

The surge in generative AI applications has led to a global arms race for compute power, positioning CoreWeave as a critical player despite its capital-intensive model.

💡 Strategic Challenges

  • CoreWeave’s CapEx far outpaces revenue, raising investor concerns over sustainability

  • CEO Intrator argues the structure is sound, with long-term contracts and secure revenue

  • Analyst Gil Luria (D.A. Davidson):

    “CoreWeave represents overflow capacity for Microsoft, which may not need that capacity in the future.”

🌐 Tariff Risk & Supply Chain Diversification

With the backdrop of U.S.-China trade tensions, CoreWeave is actively working to diversify its supply chains and mitigate tariff-related risks.

“We’re making sure our investments don’t impact our ability to protect margins,” Intrator said.

🧠 The Nvidia Connection

CoreWeave’s strategic relevance is amplified by its deep ties with Nvidia, whose chips are the backbone of its AI compute power. This positions CoreWeave as a major infrastructure-as-a-service provider in the AI revolution.

Still, the market remains cautious, with the stock reacting to the complexity of its capital structure and the uncertainty around long-term demand once client overflow needs subside.

U.S. Grid Faces Strain Amid Surge in AI Data Center Growth

/in /tarafından

The rapid expansion of AI data centers is raising concerns about the resilience of the U.S. electrical grid, with experts warning that the surge in energy demand could overwhelm the nation’s aging infrastructure. The rapid build-out of massive data centers, which can consume as much power as a mid-sized U.S. city at a single site, is driving electricity consumption to record highs. Government projections estimate that data center demand will triple in the next three years, accounting for 12% of the entire U.S. power supply.

“We are witnessing unprecedented growth, and the challenges the grid is facing are becoming more pronounced,” said Samir Vora, a senior executive at Mitsubishi Power Americas, during an interview at the CERAWeek conference in Houston.

As the demand for electricity rises, traditional fossil fuel-powered generators are being retired, and new generation and power lines are often delayed in interconnection queues, exacerbating the delicate balance required to avoid blackouts.

Mark Christie, who leads the Federal Energy Regulatory Commission (FERC), highlighted the issue at the conference, stressing that the situation has become particularly critical in the PJM Interconnection grid, which serves 13 states and the District of Columbia. This area, home to the world’s largest concentration of data centers, is also crucial for internet traffic, with Virginia alone routing 70% of global internet traffic.

In its latest capacity auction, PJM reported that prices had surged by more than 800% compared to the previous year, citing rising demand and shrinking supply. Manu Asthana, CEO of PJM, expressed cautious optimism, acknowledging that the problem is solvable, though not trivial.

PJM’s peak demand is expected to rise from 152 gigawatts to 184 gigawatts by 2030, with nearly all of the growth driven by data centers. Without substantial investments in new power supply, experts warn that these supply-demand imbalances could spread to other regions across the country, making the situation even more dire.

“It’s going to become more pronounced in other multi-state regions as well,” warned FERC’s Mark Christie, signaling growing concerns about the stability of the grid.

AMD’s AI Strategy Faces Investor Scrutiny Amid Shift to Custom Chips

/in /tarafından

Advanced Micro Devices (AMD) faces heightened investor scrutiny over its artificial intelligence (AI) strategy as Big Tech firms increasingly develop custom silicon, potentially limiting AMD’s role in AI infrastructure. The company is set to report its fourth-quarter earnings on Tuesday, with analysts forecasting a 22% revenue surge to $7.53 billion. However, competition from Nvidia and the growing adoption of proprietary chips by Microsoft, Amazon, and Meta have raised concerns about AMD’s long-term AI growth prospects.

Tech giants are ramping up investments in in-house AI chip development, benefiting companies like Broadcom and Marvell Technology, which provide hyperscalers with custom AI processors. Broadcom expects AI to represent a $90 billion revenue opportunity by 2027, a factor that helped its stock more than double last year. Marvell saw an 83% rise, while AMD’s stock fell 18% in 2024.

Despite this shift, AMD’s AI processor sales are expected to reach up to $10 billion in 2024, double its initial forecast of $5 billion. Its data center chip segment, projected to grow 82% to $4.15 billion in Q4, is expected to drive over half of total revenue. Meanwhile, its personal computer division is forecasted to rise 33% to $1.94 billion, as AMD continues to gain market share from Intel.

Supply chain constraints remain a challenge, with TSMC working to expand production capacity for AI chip packaging. However, Nvidia’s ramp-up of its latest “Blackwell” AI chips could limit AMD’s ability to secure additional manufacturing resources.

Despite these hurdles, AMD’s fourth-quarter net income is set to rise by more than 61% to $1.08 billion, reflecting strong demand for its products.