Notable tech headlines for the week: CoreWeave, TSMC, Anthropic in focus
Technology stocks slumped this week, with chipmakers and AI infrastructure providers leading losses as investors questioned the returns on heavy AI investments.
The Nasdaq Composite COMP also came under pressure, falling more than 4% for the week, while the State Street Technology Select Sector SPDR ETF (XLK) fell 7.20% over the week, reflecting broad weakness across the technology sector.
The top S&P 500 technology gainers and losers for the week are as follows:
Top Gainers:
- Fair Isaac FICO +13.91%
- Cognizant Technology Solutions CTSH +9.56%
- Tyler Technologies TYL +8.55%
- Accenture ACN +8.05%
- Gartner IT +6.66%
Top Losers:
- Ciena CIEN -25.14%
- Hewlett Packard Enterprise HPE -25.08%
- Coherent COHR -24.62%
- Super Micro Computer SMCI -24.44%
- Lumentum LITE -24.21%
Here are some of the important technology stories from this week:
CoreWeave sell-off puts AI infrastructure financing under scrutiny
CoreWeave CRWV shares plunged 12.80% on Friday as investors questioned the cost of expanding AI data centers. The sell-off spread to Nebius NBIS and IREN IREN amid concerns about borrowing costs and capital requirements.
CoreWeave's debt and lease obligations totaled approximately $52B at the end of June, highlighting the financial burden of its AI infrastructure expansion.
TSMC outlook signals resilient AI chip demand
Taiwan Semiconductor Manufacturing TSM forecasted fourth-quarter revenue of $52.3B to $54.3B, above analysts' consensus of $50.48B, providing fresh evidence of resilient semiconductor demand.
Third-quarter high-performance computing revenue rose 22% quarter over quarter. Management said AI accelerator revenue growth could exceed its previous mid-to-high-50% CAGR forecast for 2024–2029, signaling continued demand despite the broader technology sell-off.
Anthropic IPO valuation expectations cool as growth slows
Anthropic's ANTHROPIC anticipated initial public offering faces greater valuation scrutiny as investors assess its growth trajectory and AI development costs.
Reuters reported that some investors were discussing a potential valuation of around $1.5T, below earlier expectations of approximately $2T.
The AI company reportedly generated $16.8B in third-quarter revenue, up 45% sequentially, and was profitable on an adjusted basis. However, the slower growth rate disappointed some investors after revenue more than doubled from the first quarter to $11.6B in the second quarter.
The valuation discussions remain preliminary and do not represent confirmed IPO pricing.
AI spending pressures hyperscaler cash flow
Bank of America Global Research expects Microsoft MSFT to be the only major hyperscaler among its comparison group not to enter negative free cash flow territory over the next 12 months.
Amazon AMZN, Alphabet GOOGL, Oracle ORCL, and Meta Platforms META are projected to face negative free cash flow as AI infrastructure investment continues. The firm expects cash generation to recover by the second half of 2027, assuming AI revenue accelerates and capital expenditure growth slows.
Bank of America sees memory-stock sell-off diverging from fundamentals
Bank of America analysts said the recent pullback in SK Hynix SKHY and Samsung Electronics SSNLF may reflect fears of an AI spending slowdown rather than weakening memory fundamentals. Investors have worried about borrowing costs for AI cloud providers, potential Chinese supply growth, and the sustainability of demand.
BofA, however, expects memory prices, sales, and margins to remain strong through 2028. The divergence highlights uncertainty around the memory cycle: shares have come under pressure even as some forecasts point to continued demand and earnings strength.
Meta expands open AI model push with Muse Spark Open
Meta Platforms META released Muse Spark Open, an AI model designed for coding and complex tasks. It has 816B parameters, with 42B active during use, and its weights are available to developers. The model can run locally on suitably configured Apple AAPL Mac Studio computers, expanding Meta's open-model offering.