For those of you who would prefer to listen:
The Global AI Race took a surprising turn with news that a Chinese startup’s Large Language Model (LLM) outperformed the American leaders. DeepSeek’s AI model release sent shockwaves through Silicon Valley and the Market with fears that China has gained on America’s AI advantage. What’s more, DeepSeek’s LLMs were created despite U.S. aggressive export controls designed to limit China’s access to the advanced semiconductor chip that are believed to be required for AI work. Then there’s this: It’s doing it at a fraction of the cost. Simply put, DeepSeek claims to be 97% cheaper and actually performs on par with ChatGPT, Gemini, Meta AI and Claude.
The Market took a major hit Monday on this news. It clawed back most of those losses by week’s end. It was fortuitous that a number of the Tech Titans reported earnings this week, providing a transparent look into their businesses and hear their outlook and perspectives. That cooled some of the fears. But the DeepSeek impact is still felt. Like a fastball under the hitter’s chin, this was a wakeup call. Here’s another rapid rundown of what went down:
DeepSeek, a Chinese AI startup founded just last year claims it was able to train a model comparable to those from OpenAI, Google, Anthropic and Meta for a cool $5.6 Million. That sparked an immediate re-evaluation of America’s place in the AI race. Hundreds of Billions of Dollars have been spent on this innovation and much more is expected. AI has been the biggest Stock Market driver since the Dot-com days.Nvidia, considered the Godfather of AI, lost nearly $600 Billion in value on Monday. That’s the biggest single-day loss for a company in Stock Market history. Real questions are being raised as to whether the massive investments are necessary. The damage was felt far and wide.
Now I’m no AI expert. But I understand it from a Market standpoint. At issue is the notion of an open-sourced model versus closed source. There’s a big difference and neither is perfect. Open-sourced is exactly how it sounds, it’s a platform available to the masses to freely access it, manipulate it and build upon it. Open-source platforms are generally cheaper and are clearly more accessible to users. Closed source is the opposite. Its access is limited and is significantly more expensive. This has benefitted the Tech Titans who possess seemingly limitless data and sit on mountains of cash. Most can’t afford to compete. It’s a major reason these companies have dominated the Stock Market.
This is still a very fluid situation to understand the implications. Tech analysts and experts have been feverishly studying the developments. It appears to be the democratization of AI, coming faster than previously thought. Our sources think this may lead to much wider AI penetration among end users as applications can be developed with less energy and at far lower costs. Better performance with fewer resources. That’s innovation. But open source comes with its own risks. Closed source solutions are considered much safer. It’s a more controlled and safer environment. Plus, the prospects of a Chinese-led architecture are concerning. Open-sourced platforms can be used as weaponry against us. Open isn’t always better.
America has long been the most dynamic nation on Earth. Risk-taking and innovation is in her blood. Our nation has first mover advantage. It’s easier to copy success, but innovation is not in everyone’s DNA. First movers keep winning when they keep moving. It’s a race. And this AI race has massive consequences for what lies ahead.
Many of our sources question the accuracy of the reported development cost, given reports that GPU (Graphics Processing Unit) usage was significantly above what was stated. That said, with tight GPU restrictions in place with the chip embargo with China, it is difficult to track the spending. The obvious risk is the rest of the world building AI models on Chinese architecture. That would give them access to vast quantities of information. It’s very possible that the Chinese government highly subsidized this venture, making the $5.6 Million spend irrelevant. That number seems dubious at best. Besides, we can’t trust much of anything coming out of China. The way I think about it, China tells us what they want us to hear.
The White House and FBI officials are investigating whether DeepSeek circumvented the system to access Nvidia’s advanced chips. There are rumors circulating that these coveted semiconductors have been smuggled onto China’s shores by first going through Singapore. Efforts to stop the flow of Advanced AI chips to China from American companies, primarily Nvidia and Advanced Micro Devices, were spearheaded by the Commerce Department under President Biden. The approach is expected to be expanded with further tightening under President Trump.
President Trump declared he plans to ensure American leadership in AI. He indicated his administration may impose more restrictions on China. Congress is on board, with the GOP co-chair of House select committee on China urging stricter export controls on technology critical to DeepSeek’s AI infrastructure. Nvidia CEO Jensen Huang met with President Trump at the White House on Friday. Apparently, it was already on the calendar prior to the DeepSeek news. The timing is quite strategic. It sure would be fascinating to sit in on that meeting.
Concerns grew around the Globe. It was reported Italy blocked DeepSeek with immediate effect. Apple and Google already removed Deepseek from their app stores in Italy. Deepseek AI is also being questioned by Ireland’s Data Protection Commission on concerns of how it processes data of citizens. Expect more of this type of activity as clues of the mystery develop.
Trying to understand the situation and where it might be headed, I caught a fascinating interview with Dario Amodei, the CEO of San Francisco AI startup Anthropic. It owns Claude, one of the cutting edge LLM’s backed by Amazon. Amodei started as a lead engineer at OpenAI, which owns ChatGPT. Needless to say, he’s a smart guy.
Amodei had this to say about DeepSeek:
“The Chinese startup came up with software ideas to improve the efficiency of AI development and cut costs. But in the grander scheme of things, DeepSeek’s development and spending is on par with some US AI firms. DeepSeek does not do for $6 Million what cost US AI companies Billions”. He said Claude 3.5 Sonnet is a mid-sized model that cost $10-$20 Million to train. “DeepSeek produced a model close to the performance of US models 7-10 months older, for a good deal less cost, but not anywhere near the ratios people have suggested. All of this is to say that DeepSeek-V3 is not a unique breakthrough or something that fundamentally changes the economics of LLM’s; It’s an expected point on an ongoing cost reduction curve. What’s different this time is that the company that was first to demonstrate the expected cost reductions was Chinese”.
Also of great importance, there is a big difference between Large Language Models and Artificial General Intelligence (AGI). AGI will be achieved when machines become as smart or smarter than humans. Humanoids will require far more compute power which comes at a much higher price point and energy needs. Dario Amodei predicts that AGI will happen in 2026-2027 and will require “Millions of chips, and Tens of Billions of Dollars, at least.”
Amodei sees two possible scenarios:.
- In a bipolar world, the US and China will develop AGI around the same time. The “powerful AI models that will cause extremely rapid advances in science and technology” will be available to both nations. But that’s only if China gets the Millions of high-end chips needed for AGI development. Eventually, China could get an edge over the US in this scenario.
- The other scenario is a unipolar world with the US and its Western allies at the top of AI innovations. It’s unclear whether the unipolar world will last, but there’s at least the possibility that, because AI systems can eventually help make even smarter AI systems, a temporary lead could be parlayed into a durable advantage. Thus, in this world, the US and its allies might take a commanding and long-lasting lead on the global stage.
Scenario 2 is of course what Silicon Valley, Washington and the Western World are shooting for. But of course, there’s no guarantee. The stakes are high.
Back to the Market:
There was a great deal of investor complacency heading into this week. Bullish sentiment was back near highs and optimism reigned. There’s been a speculative frenzy with call options and levered ETFs, which generally occur near Market tops before corrections. The AI mania has been palpable. The Market threw $10 Trillion at AI over the last 2 years as the Magnificent 7 stocks surged. The time has been ripe for corrective price action for a while.
It just takes a spark. The asteroid hit of DeepSeek’s low-budget artificial intelligence chatbot was the cause. A Tech-wreck ensued. The big question was, can the Stock Market maintain its premium valuation if the earnings and revenue growth were about to evaporate from this Chinese threat? Fortunately, this is all happening during Earnings Season and it was some Tech Titans turn at the plate.
The 7 companies, affectionately referred to as the “Magnificent 7” were expected to report earnings growth of 21.7% in Q4. That, compared to a year ago.The remaining 493 S&P 500 companies are expected to grow earnings by a combined 9.7%. That would mark best growth since Q2 2022. Not too shabby. Looking forward, the Mag 7 earnings are expected to increase 17%+ over next four quarters, while the rest of the S&P 500 companies are expected to maintain the 9% growth rate. Expectations for broadening earnings growth is definitely key to the 2025 Bull case. The DeepSeek news certainly threatened that.
Meta announced its profits jumped 50% in 2024. The growth was driven by its massive AI bets. Management sees more, with a forecast of a “really big year.” CEO Mark Zuckerberg is sticking with its $60 Billion AI investments this year. That was key to the Market, which helped stop the slide Tuesday. Microsoft reported a double-beat led by Azure, its cloud business, which grew 31% in 2024. That’s strong growth. The problem: The Street expected 32%. The company didn’t guide for that growth to accelerate in 2025 which sent the stock lower.
Beyond the report cards, what was most critical was what the companies had to say about their outlook and the impact of DeepSeek. Both Microsoft and Meta praised the innovative Chinese AI development, but they also reiterated their aggressive AI spending plans of $80 Billion and $65 Billion, respectively. Microsoft CEO Satya Nadella said that more efficiently produced AI models will broaden adoption and fuel demand growth that will sustain the need for huge computing power. The Market loved this perspective. The Tech-trade was back on.
Meta CEO Mark Zuckerberg said his company is actually planning to incorporate DeepSeek learnings into its Llama model. He said its decision to open-source Llama in 2023 was vindicated by DeepSeek using the technology to develop its cheaper model. This allows Meta to begin reverse engineering DeepSeek’s technology as it seeks to lower the cost of training its own software. Zuckerburg asserted the more compute applied, the more intelligent the system becomes. He also doubled down on the need for more energy and infrastructure to continue the developments. The important takeaway is it is still very early days in AI. Innovation continues in a hyper competitive race. It’s very investable.
More earnings rundown:
Apple reported a double-beat. iPhone sales came in a bit light but pretty much everything else was ahead of estimates. Mac and iPad sales jumped 15%. Apple’s Services business was strong too, up 14%. Importantly, revenue from China fell 11%. Part of the issue: Apple Intelligence has yet to be approved by Chinese government. Doing business in China can be lucrative but also very difficult and costly. Another Mag 7, Tesla reported that profits fell, but the stock climbed on the back of more promise for robo taxis. Elon Musk tends to put on a show at those earnings calls.
There’s a Sleeper AI play gaining momentum. It’s known as Big Blue; That’s IBM. CEO Arvind Krishna said DeepSeek validates Big Blue’s plan. The strategy is working. IBM believes smaller, fit for purpose models will be more effective and impactful for corporate and consumer use cases. Companies are embracing the platform. Efficiency and cost are Big Blue’s advantage. The Market liked it.
Almost lost in all of this AI confusion was a Fed meeting. America’s central bank paused its rate cuts. That was expected. The debate now appears to be whether the Fed will cut at all this year. The Market pricing indicates a 90% probability that the Fed will cut at least once this year and a 62% chance that it will cut by at least a ½-point. The irony is, despite the Fed cuts, the Bond Market has sent rates materially higher. It did it almost immediately. The Bond Market is sending a message. Inflation has not been whipped, and Washington has a major spending problem. With inflation still stubbornly present, the Fed could be more Hawkish in 2025 than the Market thinks. Recalibration is taking place.
The first look at Q4 GDP, America’s Economy grew 2.3%. That was short of the 2.4% estimate and down from Q3’s 3.1% growth. So, the Economy is growing, albeit at a slower rate. But the Consumer keeps spending. Another sign of consumer strength was Visa’s earnings beat and more importantly its forecast for 2025. The company expects stronger credit card activity, already seeing an acceleration in January. Consumer spending accounts for 70% of US GDP.
There’s a major wild card hanging out there which threatens the US Economy: Tariffs. A 25% tariff hitting Canada and Mexico this weekend would increase prices and could dent economic activity. Rumors circulated that implementation would be pushed out to March. The White House quickly shut those rumors down and added that a 10% tariff on China was coming too. That sent stocks lower into the weekend. The Market doesn’t seem to believe the worst case is probable. But that could clearly change in an instant.
What seems clear: Tariffs are like a price of entry in Trump World. The problem is traditionally with tariffs, growth weakens while prices strengthen. Tariffs continue to be viewed as a tool for leverage against other countries, not a policy. The S&P remains near its all-time high. It’s clear that President Trump views the Stock Market as a scoreboard for his administration. The tariff threats have not dented the Market yet, which could further fuel his tactic.
Something else that’s been moving higher; Gold. It punched through $2800 for the first time in history. The precious metal has been sleepy strong of late suggesting the Market volatility ain’t going away anytime soon. It’s a 2025 kind of thing.
In closing, the extended Bedell Frazier family lost some cherished members. Over the Rockies, across the Great Lakes and on the other side of the fence; Mike, David and Fred made lasting marks on those they encountered. All 3 were smart, loyal and giving. They were successful professionals in their respective trades. But more importantly, they were good people who cared most about their family. They were terrific dads.
This Business of Life isn’t easy. Saying goodbye is so hard. But the beauty lies in their memories. It’s our opportunity and responsibility to keep those alive. I miss you Mike, David and Fred. You will not be forgotten.
Have a nice weekend. We’ll be back, dark and early on Monday.
Mike
