From the v4 technical report to the market narrative: why is it always TSLA that gets hurt?

$Tesla(TSLA.US) $NVIDIA(NVDA.US)

The underlying variable driving the split in the U.S. stock market in the first half of the year is hidden between two documents: one is the technical report from a large model vendor, and the other is the exchange's waiting list. The technical report determines the sustainability of the computing power narrative, while the IPO list determines the flow of money. And these two things exploded simultaneously in April this year.

First, look at the depth charge just dropped by DeepSeek.

Deepseek v4 Technical Report: https://huggingface.co/deepseek-ai/DeepSeek-V4-Pro/blob/main/DeepSeek_V4.pdf

NVIDIA, Amazon, Google, and Meta are currently enjoying the AI feast. Each has rebounded strongly from the lows at the end of March, showing great resilience because we believe in their narrative. Regardless of whether the AI "application" building is constructed, as long as someone is building, my shovels and steel are the most certain. NVIDIA has steadily risen for 11 consecutive trading days, with its total market cap reaching a high of $4.8 trillion.

Microsoft and Apple, the two unfortunate brothers, are clearly having a hard time, with year-to-date declines of about 13.6%. The worst has to be Tesla, which has fallen 12.7% year-to-date. Microsoft faces slowing cloud growth and early AI investments eating into profits; Apple is choked by the iPhone cycle rotation.

Why is Tesla always the most hurt?

First, the enemy of idealism remains that very ideal.

DeepSeek V4 has announced 1.6 trillion parameters while supporting both NVIDIA GPUs and Huawei's Ascend solution. Base model providers are focusing on inference, while Tesla's high-level autonomous driving keeps demonstrating it on the road, over and over.

Second, earnings report misjudgment.

The Q1 earnings report on April 22 actually had highlights. Adjusted earnings per share beat market expectations, revenue was about $22.4 billion, in line with market expectations. However, the pre-market stock price fell over 2% instead of rising.

Why? The officially announced capital expenditure guidance was significantly raised to over $25 billion (much higher than the previous $20 billion), with most of it going into AI autonomous driving, Robotaxi, and the Optimus robot. Musk directly told analysts that the company would spend over $25 billion on AI, robotics, and chip R&D, and free cash flow would be negative for the remaining quarters. In the short term, this is a bottomless pit.

As for who can stomp Tesla to death, NVIDIA is now an accomplice.

Because the spotlight in the second half of this year is entirely on the triple IPO concert:

· SpaceX takes the lead: listing as early as mid-to-late June. Valuation pushed above $1.75 trillion, the largest global IPO ever at $75 billion.
· OpenAI is poised: targeting a Q4 listing. The latest financing valuation has reached $852 billion.
· Anthropic is aggressive: private secondary market shares are being swapped for mansions in a scramble, with valuation expectations taking off above $1 trillion.

When three giants simultaneously raise their bowls for food, only the exchange can be charitable.

Institutions must complete one action in the second and third quarters: liquidate some of their holdings in the Magnificent Seven to free up cash to subscribe to the upcoming new shares.

Is the first target for reduced weighting the already heavily overweight NVIDIA? No. The final victim is likely TSLA, and in fact, the cut has already arrived.

Talking about bloodletting, it sounds like NVIDIA should be the first to be crossed, but the crash script might not fall on it. Why? Because NVIDIA is no longer just playing the "shovel story"; it is now the foundation for all IPO players.

Huawei is not on the field, Intel's compute cards are a joke, and AMD is still struggling with ROCm.

Where will the computing power foundation and massive costs needed for the three major AI unicorn IPOs come from? Still NVLink, H200, B100.

NVIDIA is also on the investor list for OpenAI's recent $122 billion financing round (holding about 3.47%), directly embedded in OpenAI's strategic decisions through voting rights. Regardless of their future fate, these unicorns will have to continue purchasing large amounts of NVIDIA's AI infrastructure.

Instead, those truly bleeding out in the bloodletting are the second- and third-tier growth stocks with high debt and low moats, or targets with relatively single business models (including Tesla) and weak capital drivers.

Dazi will be caught in the middle of this IPO funding war. Passive funds, in order to allocate for the giant IPOs with a valuation of $3 trillion, are collectively freeing up liquidity in the market. Even if NVDA is sold short-term, the depth of the core position for stop-loss reduction will be staggering to ordinary investors. It can withstand the fall, be the first to rebound, and even be the first to diverge.

So what was being traded in the first half of the year?

The reason for the split among the Magnificent Seven is stark: consumer-level AI monetization remains stagnant, and all expectations have shifted to the super giants, forming a new round of accumulation.

In other words, the core thread of this market in the first half of 2026 is not in fundamental updates, but in calculation:

Wait for me to cash out, then buy into the next IPO ship, so I don't lose to the market.

Before the second half's triple IPO performance begins, there is only one golden rule to hold this card game: those alive in the position will always end up with more than those who just watch the highs and lows.

The entire market is extremely volatile, either pulling or smashing. The smashing even has its own logic: smash → money into money market funds → feed the giants → wait for IPO.

This closed loop ran through again a few days ago. The last time it was smashed was non-farm payrolls; this time, it's Iran. No need to guess whether next month's non-farm payrolls will be 170k or 180k, or whether the US and Iran will fight or make peace. Just look at one thing: how much was pulled into money market funds today? $7.8 trillion, and still rising.

This is the main trading theme of the first half. Technical reports and IPO prospectuses have more pricing power than any macroeconomic data.

DeepSeek V4 writing Ascend into its official documentation reminds everyone that the computing power cost curve is being braked by a new player. When inference efficiency becomes the only hard currency in the next phase, whoever can run more business scenarios under the same computing power budget will survive the tide of IPO bloodletting.

Tesla chooses to build its autonomous driving moat with road test data, a path destined to burn money, time, and patience. But on the other hand, those who truly understand the long-term flywheel won't lose their chips in the short-term pain of capital expenditure.