capex

AI compute

2026-07-03 00:01:11

Meta's idle GPU rental, Palantir's token outrage, and the rise of Chinese open-source models: AI CapEx shifts from total quantity to structural stratification

Meta's potential plan to rent out excess AI compute triggered a market rout. Palantir CEO Alex Karp slammed the token-based billing model as a 'wealth tax.' This article dissects the structural shift in AI compute spending: cloud giants raise prices while model companies face utilization pressure; Chinese open-source models like GLM-5.2 become enterprise cost-cutting tools; token consumption explodes but CFOs scrutinize every unit. Coinbase halved AI spending through model routing. The market enters a phase where compute is stratified: top-tier remains scarce, mid-tier becomes awkward, and bottom-tier faces compression.

1400

AI compute

2026-07-02 19:01:07

Meta Considers Selling Spare Compute, Palantir Slams Token Tax, Zhipu Emerges as Silicon Valley Darling: AI Capex Enters the Era of Compute Stratification

Meta's move to potentially sell unused AI compute spooks the market, while Palantir CEO Alex Karp condemns OpenAI/Anthropic's token pricing as a 'wealth tax' on enterprises. Meanwhile, Chinese open-source models like GLM-5.2 gain traction in Silicon Valley, with Coinbase slashing AI costs by nearly half after switching. This article dissects the shift from a 'total AI capex' story to a 'structural' one: top-tier deterministic compute remains scarce and expensive (AWS raised prices twice), middle-tier compute faces utilization scrutiny, and bottom-tier is squeezed by open-source alternatives. Data shows cloud revenue still surging (AWS Q1 +28% to $37.6B), but 60% of enterprises are capping token spend post-pilot. The real debate is no longer about building the most machines, but running them profitably.

1300

TSMC

2026-07-01 04:31:07

TSMC Q2 Earnings Preview: Can ~66% Gross Margin Hold? AI and 2nm as Key Variables

TSMC will report Q2 2026 earnings on July 16, with market focus on three core questions: the strength of AI orders, the ramp-up of 2nm process, and the sustainability of gross margin near 66%. Q1 revenue reached $35.9B, gross margin 66.2%, EPS NT$22.08. Q2 guidance calls for revenue $39B–$40.2B and gross margin 65.5%–67.5%. AI/HPC remains the strongest growth driver, with HPC contributing 61% of Q1 revenue and advanced nodes 74% of wafer revenue. However, capex is climbing to $52B–$56B, and overseas fabs add cost pressures. N2 entered high-volume production in Q4 2025, with capacity CAGR expected at ~70% from 2026 to 2028. Investors should watch management’s tone on order visibility, inventory, and margin outlook.

1520

TSMC

2026-06-30 23:31:02

TSMC Q2 Earnings Preview: Can 66% Gross Margin Hold? AI Demand and 2nm Ramp-Up in Focus

TSMC will report its Q2 2026 earnings on July 16, with market attention shifting from revenue growth to three key issues: whether AI orders remain strong, whether the 2nm ramp is on track, and whether gross margins near 66% can be sustained. After a strong Q1 (USD revenue $35.9B, gross margin 66.2%), Q2 guidance stands at $39.0-$40.2B with gross margin of 65.5%-67.5%. Morgan Stanley estimates Q2 gross margin at 67.4%, the upper end of guidance. AI/HPC continues to drive revenue structure shift, but high capex ($52-$56B) and 2nm ramp costs pressure margins. This article provides a comprehensive preview, analyzing AI orders, advanced packaging, 2nm progress, and indirect implications for crypto mining and AI token markets.

1450

TSMC

2026-06-30 22:31:10

TSMC Q2 Earnings Preview: Can 66% Gross Margin Hold Amid AI Demand and 2nm Ramp-Up?

TSMC will hold its Q2 2026 earnings call on July 16 at 14:00 Taiwan time, with the market focusing on three key questions: whether AI orders remain robust, whether the 2nm process ramp-up is on track, and whether its near-66% gross margin is sustainable. In Q1, TSMC reported USD revenue of $35.9 billion and gross margin of 66.2%. Its Q2 guidance calls for gross margin of 65.5%-67.5%. Morgan Stanley expects Q2 gross margin to reach 67.4%, near the high end of guidance. AI/HPC demand is the main growth driver, but costs from advanced node expansion, depreciation, and yield learning could pressure margins. TSMC has guided full-year 2026 CapEx to the high end of $52-$56 billion, indicating long-term capacity needs. The 2nm ramp, overseas fabs, and management's commentary on customer orders, inventory, and cost pressure will be critical for the stock's next move. Investors should watch the July 16 conference call for signals on H2 demand, pricing power, and margin outlook.

1400

TSMC

2026-06-30 15:30:59

TSMC Q2 Earnings Preview: Can the 66% Gross Margin Hold? AI Demand, 2nm Ramp-Up, and CapEx in Focus

TSMC will announce its Q2 2026 earnings on July 16, 2026, with market focus on three key variables: the sustainability of robust AI/HPC demand, the progress of 2nm process ramp-up, and whether gross margin can stay near 66%. In Q1, TSMC reported revenue of $35.9 billion, gross margin of 66.2%, and operating margin of 58.1%. Q2 guidance calls for revenue between $39.0B and $40.2B, with gross margin of 65.5%-67.5%. Morgan Stanley expects gross margin to reach 67.4%. CapEx guidance points to the high end of $52B-$56B for 2026, with 2nm capacity CAGR of ~70% from 2026 to 2028. The earnings call will be a critical test of the AI hardware supply chain's profitability.

1420

TSMC

2026-06-30 13:01:01

TSMC Q2 Earnings Preview: AI Demand Supports 66% Gross Margin, 2nm Ramp Becomes Long-term Key Variable

TSMC will report Q2 2026 earnings on July 16. The market is focused on three key issues: whether AI/HPC demand remains robust, the progress of 2nm mass production ramp, and whether the gross margin can stay above 65%. In Q1, gross margin reached 66.2% with revenue up 35.1% YoY. Morgan Stanley expects Q2 gross margin of 67.4%. Full-year capex is guided to the high end of $52-56 billion, signaling management's confidence in long-term capacity gaps. However, depreciation from advanced node expansion and overseas fabs may gradually pressure margins. TSMC's strength hinges on the AI hardware chain—if any link (demand, yield, costs) disappoints, the near-66% gross margin will be tested.

1450