Seeking Deeper Meaning in AI
DeepSeek was the big story in markets this week with its low-cost, open-source AI model, sparking discussion about the extent of U.S. technology spending. It also raised broader questions about the U.S. lead in the global AI race. Microsoft# and OpenAI are apparently investigating whether DeepSeek improperly obtained OpenAI data to build its model. Not to be outdone, Alibaba#, another Chinese company, announced that its latest AI model update can outperform DeepSeek along with OpenAI, Meta’s# Llama, and Google’s# Gemini. A lot of claims are being made that need to be backed up, but for now the uncertainty puts a short-term cloud over the AI space.
DeepSeek’s “breakthrough” challenges assumptions about the cost of investing in AI and heightens scrutiny on the outsized capital spending ramp. Open AI and Nvidia both praised DeepSeek’s results and both OpenAI and Meta are looking to better understand DeepSeek’s methods. There are skeptics who doubt its cost claims and the app’s capabilities (many were unable to register) and whether that prompts the US to tighten China export controls further. The Chinese government’s influence and censorship are other factors seen limiting DeepSeek’s adoption outside of China, and the app could potentially be banned similar to Tik Tok’s pending ban.
The impact remains to be seen, though hyperscalers like Microsoft and Meta may re-evaluate the massive level of spending on computers and equipment to generate and train AI language models. Just last week, Microsoft’s close partner OpenAI announced Stargate, a joint venture with SoftBank and Oracle that intends to spend up to $500 billion on new data centers over several years. These companies will probably also focus on more efficient methods to achieve their performance goals after DeepSeek’s recent announcements. Either way, it is a bit of a wakeup call for industry leaders.
Microsoft reported results last night ahead of expectations, saw decent growth in its AI-related business, and posted 32% growth in its Azure cloud-computing revenue. Meta also posted better than expected revenue and earnings last night. Both companies expect to continue to spend tens of billions of dollars on AI-infrastructure in 2025. ASML# noted solid orders on Wednesday for its semiconductor chip-making equipment. Lam Research# also reported earnings ahead of expectations last night with an outlook above expectations for next quarter for its chip-fabrication equipment. Near term data points are positive, but the question remains what impact a more efficient and low-cost AI model means. Even power producing companies got caught up in the downdraft this week, with the assumption that less power use for AI training and inference could mean less growth for electricity and all forms of its production.
Steady at the Fed
The January Federal Reserve meeting ended yesterday with no change to the Fed’s benchmark rate at 4.25%-4.50%. This was broadly expected, and follows three straight cuts since September 2024 worth a full percentage point. The Fed’s statement pointed out that labor market conditions remain solid, though inflation still remains somewhat elevated. The statement also noted the range of possibilities is very wide for the impact of tariffs. Fed Chairman Powell again assumed a data-dependent posture, as usual, and noted that the Fed is not in a hurry to lower interest rates further. Expectations are currently for a 70% chance that the Fed holds rates steady at its next meeting in March.
Grounds for Dismissal
Hitting much closer to home, coffee prices set a record on the recent trade spat with Columbia. Arabica coffee prices hit a record level as traders digested President Trump’s threats to impose tariffs and economic sanctions on Colombia. The U.S. imports about 30% of its coffee from Colombia. Trump had issued threats to impose 25% tariffs and sanctions in retaliation for the Colombian government refusing to allow two military planes with migrants to land. He changed course later as Colombia agreed to repatriate migrants into the country. Connecting the dots, Starbucks# recently reported that its prices were up 3% over the prior year, but transactions were down 8%. Higher prices may be one reason driving traffic numbers down, but a turnaround is being attempted at the company. Perhaps some of the tariff threats will turn out to be more of a negotiating tactic and transactional tool and won’t amount to a hill of beans in the long run.
Actor Christian Bale turns 51, singer Phil Collins turns 74, and Dick Cheney is in the hunt for 84.
Christopher Crooks, CFA®, CFP® 610-260-2219