The Chinese firm DeepSeek just upset the computing and investing world that relies on a smarter approach, rather than using more expensive computing power. Currently, Artificial Intelligence relies heavily on Graphic Processing Units (GPU’s) that cost $40,000 each and need massive data centers that require thousands of GPUs. The energy needed to run a data center has led Tech behemoths to built power plants adjacent to the data center. By 2030 data centers are forecast to consume 11% - 12% of total US power demand from 3.5% currently, according to McKinsey and Company. DeepSeek presented an approach that allegedly lowers the cost from $100 million to $5 million and still provides the same performance. Understandably there are doubters, but DeepSeek is an open source application that allows anyone access to the methodology and results. This will allow other firms to duplicate or find short comings.
DeepSeek suggests that those pursuing Artificial Intelligence won’t have to spend top dollar on Nvidia GPU’s or employ a small army of programmers to achieve a similar result. Nvidia’s stock plunged -17% wiping out $560 billion in market cap. The losses extended to other Semiconductor stocks which caused the Semiconductor ETF SMH to lose -9.8%. The shock wave spread to Utility stocks as investors reassessed the need for new power plants to feed the data center energy demands. Stocks on the doorstep of developing Small Modular Reactors that were expected to provide nuclear power for data centers were hit too. The Uranium ETF (URA) dropped -11.2%, as the need for more power generation evaporated with the prospect of cheaper less power hungry GPU’s.
Microsoft’s CEO Satya Nadella posted on Monday about the “Jevons paradox”. If a new approach lowers the cost of entry for Artificial Intelligence more firms will be able to use it and the demand for AI will increase. The bad news in the short term will be offset over time as future demand makes up for the loss of revenue. The problem in the short term is Nvidia and other Mag 7 stocks are priced for little or no competition, which is why investors were willing to pay up. As the cost of hardware (GPU’s, chips) declines, the producers of chips will suffer, but software firms that use hardware will benefit as their cost of doing business falls. This is why Meta’s stock gained 1.9%, while Nvidia lost -17.0% and Broadcom shed -17.44.
Spending on AI by the Mag 7 stocks will likely be maintained for awhile as each company determines the longer term impact of this development. This approach will convey to investors that the confidence the big spenders have in AI hasn’t been damaged. It also represents a hedge on the future. If DeepSeek’s claims are overstated, the big firms will preserve their advantage. My guess is that spending on AI hardware will be scaled back at a minimum and could actually decline by the end of 2025.
The Trailing P/E for the bottom 493 stocks in the S&P 500 is 19.8, while the P/E for Broadcom is 66.5, Nvidia 63.4, Amazon 40.9, Microsoft 33.6, and Meta 25.7. The average P/E for these 5 stocks is 46.0 or 2.3 times the P/E for the bottom 493 stocks. These numbers changed today.
The FOMC will leave the Funds rate unchanged at their January 29 meeting which won’t be a surprise. Wall Street will focus on Chair Powell’s comments during the press conference to discern whether the FOMC might lower the Funds rate at the March 19 meeting. Chair Powell won’t tip his hand and will emphasize how data dependent (employment and inflation) the FOMC is. Chair Powell won’t express an opinion about whether President Trump’s tariffs and deportation policies will cause inflation to increase since the details are unknown. But the consensus of FOMC members about the policies was made clear at the December FOMC meeting. The majority of FOMC members expect inflation to move higher as a consequence of President Trump’s trade and immigration policies. In the December Summary of Economic Projections (SEP), the Median for the Headline PCE was increased from 2.1% in September to 2.5% in the December SEP, and the Core was lifted to 2.5% from 2.2%. These are significant increases and reflect the level of concern most FOMC members feel.
The December Personal Consumption Expenditures Index (PCE) will be released on January 31. In November the Headline PCE was up 0.13% and the Core increased by 0.11%. There is a good chance the December Headline and Core PCE will increase by 0.2%.
Stocks
My guess is that the DeepSeek news will convince some long time holders of Nvidia and Semiconductor stocks to lower their overweight positions by selling into rallies, until it can be determined whether a seminal event has occurred. In the 1980’s Mainframe computers were the kingpins of large computing, but the advent of personal computers introduced a change that took almost two decades to be realized as the cost of computing collapsed. By the mid 1980’s the average personal computer had far more power than the computer on board the Apollo 11, and today the average personal phone has more than 100,000 times more computing power. The announcement by DeepSeek is probably a game changer in the world of Artificial Intelligence but it will take time to alter the real world. The outcome will bring the cost of AI down so much that it will become a commodity. IBM’s initial personal computer cost more than $4,000 and came with 64K of memory and floppy discs!
The large decline in Nvidia and other Technology stocks ignited a rotation into stocks that have been lagging. Although the S&P 500 declined -1.46% more stocks went up (1604) than down (1203). Most mutual funds and pension funds are 95% invested all the time. As they sold the tech stocks they used the proceeds to buy sectors deemed immune from the tech carnage. Financials, Health Care, and Consumer Staples were the biggest winners.
Nvidia’s stock experienced a Key Reversal on January 7 and the decline on January 27 blew through the support near $130.00 (red trend line). Intraday Nvidia tested the green line of support just below $120. Ultimately, Nvidia is expected to test the blue line of support just below $100 after any near term bounce.
The Semiconductor ETF (SMH) closed above the red trend line for the first time since the low in August on January 22. The breakout was reversed on January 24. The green support line was tested on January 27. The expectation is that the blue support line below 200 will be tested after any near term bounce.
Last week I noted that the rally off the low of 5773 in the S&P 500 was 5 waves which implied that the S&P 500 would rally to a new high above 6100, after a pullback. The pullback never materialized and instead the S&P 500 went to a new high directly. The decline on January 27 created options. The S&P 500 gapped lower on January 27 and left an Island Reversal after the S&P 500 gapped higher on January 22. This is a negative chart pattern since it suggests the gap up and gap down represent a sign of exhaustion. It is noteworthy that the S&P 500 created gaps on January 15, January 17, and January 22. The frequency of so many gaps indicates emotional buying which is not positive. The S&P 500 closed the gap from January 17 when it traded down to 5963. The gap from January 15 is at 5871 so a further drop to close this gap must be respected.
Despite the gap lower on January 27 it was positive that more stocks went up than down. This was a knee jerk response to the big hit in the Tech stocks. If breadth continues to hold up as Tech stocks experience more selling that brings the S&P 500 down to 5871, it will increase the odds that the rally to 6128 was Wave 1 and this pullback is Wave 2 of Wave 5. This would set the S&P 500 up for another rally to a higher high in Wave 3 of Wave 5. A close below 5773 would eliminate the bullish potential.
As the S&P 500 rallied to 6128 the NYSE Advance – Decline Line was -3,096 issues below the peak on November 29. This is a sizable Negative Divergence. In 2007 the A-D line’s negative divergence was -986 issues when the S&P 500 recorded a new high on October 9. On January 4, 2022 the A-D Line negative divergence was -3840 when the S&P 500 reached a new all time high. A Negative Divergence is one of the more reliable Technical signals since it underscores that the broad market is weaker than the new high in the S&P 500 suggests. If the economy is not in a recession, the Advance – Decline Line Negative Divergence has frequently presaged declines of -10% to -15%. If the economy does slip into a recession, the A-D Line Negative Divergence has preceded bear market declines of -25% to -50%, especially if the S&P 500’s valuation is as stretched as it is now.
The sharp sell off on January 27 was concentrated in Semiconductors and Technology stocks, which isn’t a reason to sell all stocks since it doesn’t reflect broad economic weakness. This is why it’s important that market breadth and the Advance – Decline Line hold up if there is additional selling in Semiconductor and Tech stocks.
Dollar
After reaching the high of 110.17 the Dollar fell to 106.96 after declining in 5 waves. In the process the Dollar closed below 107.50 (red horizontal line). As I noted in the last three WTR’s a close below 107.50 would confirm a top. After completing 5 waves down, the Dollar is set up to bounce. A rebound to 108.50 – 108.75 would represent wave 2.
The drop from 110-17 to 106.96 is the first wave of a larger decline that is expected to unfold throughout 2025. From a high of 114.74 in September 2022 the Dollar declined 15.21 points to 99.59 in July 2023. An equal decline of 15.21 from 110.17 would bring the Dollar to 95.00. The first decline from 114.74 lasted from September 26, 2022 to July 10, 2023 or 41 weeks. If the coming drop is equal in time a low would develop in late October or early November.
At some point in 2025 President Trump or Treasury Secretary Bessent will note that a strong Dollar is not good for the profits of US international companies, or the goal of increasing exports since a strong Dollar makes US exports more expensive, and it makes it tougher to increase manufacturing jobs in the US.
Treasury yields
Inflation concerns are overblown if President Trump’s tariff and deportation plans wind up being less onerous than markets are bracing for. This will create the opportunity for a positive surprise. However, inflation will have to ease and the economy will need to slow for a sustained rally to take hold. I still think that’s the more likely outcome. The FOMC won’t be able to lower the Funds rate until they can asses the inflationary impact from tariffs and deportations, and inflation will need to fall. The December PCE will likely show that monthly inflation ticked higher from the low numbers in November.
TLT has closed above the blue trend line at 87 and tested the red horizontal trend line is at 88.25. This represents important resistance as it marks the last high before the breakdown below the blue trend line at 87.00. The rally from 84.89 is so far a corrective 3 waves, which opens the door for another lower low.
At the last two important trading lows (October 2023, April 2024) TLT recorded a low, bounced and then dropped to a lower low. This allowed the RSI to record a positive divergence as the RSI made a higher low (green line on bottom panel for RSI). TLT has bounced and a positive divergence will develop if TLT closes below 85.43.
Gold
The rally since Gold fell to 2541 on November 14 has been so choppy that it looks like a corrective bounce within the context of a larger correction. On January 24 Gold traded up to 2785. I think this rally is the end of the Wave B (blue) that started after Gold dropped to 2541, even if Gold briefly trades above 2789. If correct Gold will eventually trade below 2541 before the Wave 4 correction from the September 2022 low is complete. Gold rallied from the September 2022 low of 1812 to 2789 in Wave 3. If Gold corrects 38.2% of the $977 rally Gold could fall to 2415 before Wave 4 is complete. Once the Wave 4 correction is complete Gold is expected to rally above 3000.
Gold Stocks
After trading down to 33.42 on December 30 GDX rallied to 38.16. The rally has been choppy which suggests that it is a counter trend move. As long as GDX doesn’t trade above 39.17 a decline below 33.42 is expected.
After bottoming in February 2024 GDX rallied from 25.67 to 44.22. The 61.8% retracement of the $11.46 rally would bring GDX down to 32.75. GDX declined from $44.22 to $35.19 ($9.03), and an equal drop from $39.17 would bring GDX down to $30.14. If Gold does drop below $2541, GDX could easily fall to near $30.00. GDX is expected to trade below $33.42, at a minimum.
Major Trend Indicator
The MTI closed below the red moving average on February 21, 2023 which represented a short term sell signal with the S&P 500 at 3997. On March 28, 2023 the MTI generated a short term buy signal when it crossed above the red moving average with the S&P 500 at 3971. On August 3, 2023 the S&P 500 closed at 4502 and the MTI provided a short term sell signal. The MTI generated a short term buy signal on November 2, 2023 when the S&P 500 was 4318. A short term sell signal was generated on January 5, 2024 when the S&P 500 closed at 4697. This signal was reversed on January 19 when the S&P 500 closed at 4840. The MTI closed below the red moving average on March 18 after the S&P 500 closed at 5149, but reversed on March 21 when the S&P 500 closed at 5241. As I noted at that time, “Given bullish sentiment and the S&P 500’s price pattern, this short term buy is likely to be a whipsaw.” The MTI closed below the red moving average on April 2 when the S&P 500 closed at 5211. On April 5, and for the first time since November 13 when the S&P 500 closed at 4411, the MTI closed below the green moving average with the S&P 500 closing at 5204.
On March 28 the daily value of the MTI increased on May 6 with the S&P 500 at 5181. On May 8 the MTI closed above the red moving average, triggering a short term buy signal by the MTI with the S&P 500 at 5188. The S&P 500 has rallied to a new all time high and the Advance – Decline Line confirmed the new high on July 12. Update: On July 19 the MTI closed below the red moving average reversing the May 8 buy signal. The S&P 500 closed at 5505 on July 19.
On March 28 the daily value of the MTI increased on May 6 with the S&P 500 at 5181. On May 8 the MTI closed above the red moving average, triggering a short term buy signal by the MTI with the S&P 500 at 5188. The S&P 500 has rallied to a new all time high and the Advance – Decline Line confirmed the new high on July 12. Update: On July 19 the MTI closed below the red moving average reversing the May 8 buy signal. The S&P 500 closed at 5505 on July 19.
The MIT bottomed on September 12 and crossed the red moving average on September 16 when the S&P 500 closed at 5633. The rally on October 11 in the S&P 500 prevented the MTI from closing below its red moving average. The MTI crossed below the red moving average on October 25 when the S&P 500 closed at 5808.12. The big rally after the election lifted the MTI above the red moving average on November 7 when the S&P closed at 5973. Update: The MTI crossed below the red moving average on December 16 when the S&P 500 closed at 6074.08
The MTI peaked at a high level in March 2024 when the S&P 500 was trading at 5265, and reached a lower high although still strong in July 2024. The MTI recorded another more pronounced lower high in December when the S&P 500 traded up to 6100. The MTI declined below the blue horizontal on January 13 which is another sign of weakness. When the S&P 500 recorded a new all time high on January 24 the MTI was far below its high in December just 6 weeks ago. This is a warning of weakness and loss of upside momentum. The magnitude of divergence with the December high is a warning that a large decline is becoming more likely if a good reason to sell materializes in coming months.
Another signal that an important top may have been signaled is coming from the Advance – Decline Line. On January 24 the S&P 500 recorded a new all time of 6128 while the NYSE Advance – Decline Line was -3,096 issues below its peak on November 29. This is a sizable Negative Divergence. In 2007 the A-D line’s negative divergence was -986 issues when the S&P 500 recorded a new high on October 9. The S&P 500 subsequently fell by -57% after a really good reason to sell developed (financial crisis). On January 4, 2022 the A-D Line negative divergence was -3840 when the S&P 500 reached a new all time high. As the FOMC hiked rates and provided a good reason to sell, the S&P 500 dropped -28%.
A Negative Divergence is one of the more reliable Technical signals since it underscores that the broad market is weaker than the new high in the S&P 500 suggests. If the economy is not in a recession, the Advance – Decline Line Negative Divergence has frequently presaged declines of -10% to -15%. If the economy does slip into a recession, the A-D Line Negative Divergence has preceded bear market declines of -25% to -50%, especially if the S&P 500 valuation was stretched as it is now.
The Daily Shot A number of charts in this letter were from The Daily Shot.