CoreWeave, Lumentum, and Sandisk were among the worst-performing stocks in the Nasdaq 100 last week. On Monday, Sandisk and Lumentum rebounded, while CoreWeave slipped further.

CoreWeave was at the heart of the brutal selloff among artificial-intelligence-related stocks, plunging 25% last week as fears grew over the financing needed for the massive AI-infrastructure buildout.

The cloud-computing company, which pioneered the so-called neocloud sector of selling AI computing, saw its corporate bonds sell off, signaling credit investors' concerns about the company's mounting debt and future borrowing costs.

The selloff contributed to a tough week for the broader market. The five worst-performing Nasdaq 100 stocks were Rocket Lab, Lumentum, CoreWeave, Nebius, and Teradyne.

Sandisk stock also tumbled about 18% last week, placing it among the 10 worst-performing Nasdaq 100 components.

But with the start of a new week, investor sentiment showed signs of improving.

Sandisk stock rose 6.1% and was among the best performers in the Nasdaq 100 and the S&P 500. Rocket Lab rose 2%, while Lumentum stock advanced 2.4%, and Nebius gained 1.5%. Teradyne rose 3.6%.

CoreWeave bucked the trend.

CoreWeave fell 2.1% to $60.06 on Monday after rising as high as $65.92 intraday. The decline came even as Goldman Sachs analyst Gabriela Borges upgraded the stock to Buy from Neutral.

The five worst-performing stocks in the S&P 500 last week were Ciena, Coherent, Lumentum, Hewlett Packard Enterprise, and Skydance. Much like the Nasdaq 100 stocks, many of the S&P 500 laggards bounced back Monday.

Ciena rose 4.1%, Coherent gained about 0.1%, and HPE stock advanced 1.5%. Shares of Skydance, however, declined 3.8%.

More broadly, the iShares Semiconductor ETF, or SOXX, rose 1.9% on Monday and was on pace to snap a nine-session losing streak.

William Blair analyst Richard de Chazal said in a research note Monday that financial markets last week revealed that even with solid demand, the AI buildout now depends much more on credit markets than was previously thought.

"In our minds, it is like a landlord borrowing heavily to build an office block, with three anchor tenants taking most of the space. The leases look secure, but if one tenant moves out at renewal or demands a lower rent, the landlord is still on the hook for the mortgage. As a result, investors are now demanding extra compensation for taking on that risk, meaning the pace of the buildout outside the hyperscalers is likely to slow unless borrowing costs come down," de Chazal said.

Wall Street seems to be hoping the first batch of Big Tech earnings this week can help shift the AI narrative back toward demand for computing power and capital spending, rather than concerns about how that spending will be financed.

Monday's rebound offered some relief, but upcoming earnings will test whether it can last.