The Nvidia Hugging Face acquisition, valued at $12.9 billion in total, dominated portfolio thinking last week as equity markets opened September on uncertain ground, with rising oil prices and Treasury yields testing investor appetite for the AI buildout’s biggest beneficiaries.
For the week, the S&P 500 eked out a 0.1% gain while the Nasdaq gained 0.4%, recovering from a rough start to what is historically a difficult month. Both indices recorded their fifth weekly gain from the past six. A stronger-than-expected August jobs report pushed bond yields sharply higher on Friday, with the 2-year Treasury yield reaching its highest level since January 2025. That data added to inflation concerns already stoked by rising oil prices amid renewed fighting between the US and Iran. Market odds of a September Federal Reserve interest rate hike rose to 58%, up from 49.4% a day earlier, according to the CME FedWatch tool.
Rebalancing away from the AI buildout
Against that backdrop, the portfolio made several meaningful adjustments, trimming exposure to some of the AI buildout’s biggest winners while rotating into more defensive names. Even Nvidia’s blowout earnings the prior week failed to spark a broader AI rally, reinforcing the concern that investors are becoming less willing to reward even strong results across the group.
Palo Alto Networks was trimmed ahead of its earnings, locking in a roughly 148% gain on shares purchased in August 2024, following a run of more than 80% this year. Corning was exited on Tuesday, locking in an average 52% gain, even though shares remain up 70% this year, they have dropped more than 40% from record highs in late June. Capital from those sales was redeployed into BNY and Kimberly-Clark, each starting at roughly 1% weightings in the portfolio. Kimberly-Clark carries a 4.75% dividend yield and a potential catalyst from its pending Kenvue acquisition. Micron was also added to, bringing that position to about 1%, with AI-driven memory demand continuing to outstrip supply.
Palo Alto Networks did deliver strong results. CEO Nikesh Arora said roughly $1 trillion of global cybersecurity infrastructure needs to be modernised to defend against emerging threats. ‘You cannot deploy AI successfully if you do not get cybersecurity right,’ he said on CNBC’s ‘Mad Money’. The stock was upgraded to a hold-equivalent 2 rating from a sell-on-strength 3 following Wednesday’s pullback, with the price target raised to $400 from $380. Shares still sold off on profit-taking, leaving Palo Alto as the portfolio’s worst performer for the week, losing more than 10%.
Broadcom presented a separate test of the market’s patience. Despite better-than-expected revenue and earnings and a bullish long-term AI outlook, the stock ended the week down 3%. CEO Hock Tan raised his fiscal 2027 AI revenue forecast to $115 billion and expects it to double to $230 billion in fiscal 2028. Concerns around customer concentration, vendor financing, and data centre pushback continued to weigh. The price target was lowered to $430 from $480, with the hold-equivalent 2 rating maintained. ‘The only regret I have is … I didn’t just get rid of it all,’ Jim Cramer said during Thursday’s Morning Meeting, though he added he is willing to hold the remaining position because of Broadcom’s deepening relationship with AI startup Anthropic, which is set to go public this fall.
Nvidia Hugging Face acquisition: deal structure and strategic logic
The week’s most structurally consequential development was Nvidia’s $12.9 billion move on Hugging Face, an open-source AI platform used by more than 18 million developers. According to Winzheng, the deal comprises approximately $11.9 billion in cash paid to shareholders, with an additional $1 billion in equity set aside to retain core employees, a structure that signals Nvidia’s intent to keep the platform’s talent in place rather than simply absorb its user base.
The strategic rationale parallels Microsoft’s 2018 purchase of GitHub. Rather than directly monetising the platform, Nvidia can deepen developer loyalty, make its hardware and software the natural choice for deploying AI models, and keep a strategically valuable asset away from competitors. The acquisition also hedges against the risk that hyperscalers accelerate their own custom chip programmes: whether developers use open-source or proprietary models, both require massive amounts of compute.
Context on scale matters here. AsiaConnect Magazine reports that Hugging Face’s annualised revenue stands at around $150 million, meaning Nvidia is paying a steep multiple for ecosystem positioning and developer mindshare rather than near-term cash generation.
Nvidia itself is up roughly 22% this year and trades at just 14 times next year’s expected earnings, a level Jim Cramer described as too cheap given the company’s growth and expanding AI ecosystem. He noted that a larger share buyback could provide an additional catalyst; the stock rose 6% last week, edging closer to its mid-May record highs.