Two days ago, Michael Burry called the semiconductor rally a replay of 1999. On Wednesday, SoftBank reported annual profits that more than quadrupled to $32 billion, fueled by a $45 billion gain on its OpenAI stake. The AI trade's defining tension landed in a single week.
VonTrend is a financial media publication for informational purposes only. We are not financial advisors. This may contain paid advertisements and affiliate links for which we may receive compensation. Nothing on our website should be considered personalized investment advice. Always consult a licensed financial professional before making investment decisions.
SoftBank's fiscal year numbers are staggering. Net profit hit 5 trillion yen, roughly $32 billion. That is more than four times the prior year.
The driver was OpenAI. SoftBank invested $34.6 billion into the ChatGPT maker over 12 months. The fair value of that stake reached $79.6 billion by the end of March. That is a $45 billion gain on a single position.
The company's net asset value hit a record 40.1 trillion yen, roughly $250 billion. It has since climbed above $300 billion. SoftBank now plans to invest another $20 billion in OpenAI through two tranches in July and October, bringing its cumulative bet to $64.4 billion for roughly a 13% stake.
One number in the report deserves attention. SoftBank downsized a margin loan backed by its OpenAI stake from $10 billion to $6 billion.
That means lenders, even while watching SoftBank book record profits, trimmed their own exposure to the same AI thesis driving those gains. When the people lending the money want less risk than the people investing it, that is worth noting.
On Sunday, VonTrend published Michael Burry's position. The investor who predicted the 2008 housing crash bought January 2027 puts on the SOXX semiconductor ETF, Nvidia, Palantir, and Oracle. He compared the current rally to the final months before the 2000 crash.
Palantir fell 4.3% on Wednesday. But SOXX and the broader semiconductor index rallied, pushed higher by the news that Nvidia's Jensen Huang joined the presidential delegation to Beijing.
The Burry thesis is not dead. It is being tested against an AI trade that keeps producing numbers too large to dismiss.
Cerebras is expected to price its IPO on Thursday at or above the $150 to $160 range. At the high end, the company would raise $4.8 billion at a valuation near $49 billion. Demand is 20 times the available shares.
What makes Cerebras different: it is profitable. The company reported $510 million in revenue and $238 million in net income. Most AI chipmakers going public burn cash. Cerebras makes money.
Fervo Energy, a geothermal company backed by Bill Gates, gave another signal on Wednesday. It priced its IPO at $27 and indicated an open near $35, a 30% first-day pop. The IPO window is wide open.
Applied Materials releases fiscal second-quarter results after the bell on Thursday. Wall Street expects $7.69 billion in revenue and $2.68 in earnings per share, roughly 12% growth year over year.
This is the semiconductor equipment bellwether. When companies like Applied Materials beat and raise guidance, it confirms chipmakers are still spending aggressively on new capacity. When they miss, the bear thesis gains teeth.
Morgan Stanley raised its price target to $454 this week. RBC forecasts a slight beat with a strong forward outlook.
The bull case produced hard numbers this week. SoftBank's $45 billion gain is not a projection. It is a mark-to-market on an investment that more than doubled in 12 months. Cerebras is about to go public at a profitable $49 billion valuation. Nebius Group surged 17% on Wednesday after reporting revenue that grew eightfold.
The bear case also has data. The Shiller CAPE ratio sits at 40. SOXX is up more than 60% this year. Rate hike odds crossed 50%. Producer inflation hit 6%.
Both sides are arguing with real numbers. The question is not whether AI is real. It is whether the market has priced in too much of it, too fast. Applied Materials and Cerebras will add two more data points by Friday.