The most important signal today was not the Nasdaq rising roughly 2%. It was money moving back into the entire AI supply chain.
With $苹果(AAPL)$ as the only exception, the Magnificent Seven rallied almost across the board. $Meta Platforms, Inc.(META)$gained roughly 6%, $微软(MSFT)$ and $谷歌(GOOG)$nearly 5%, Amazon 4.6%, Tesla 3.5%, and Nvidia around 3%. Amazon’s market capitalization also crossed $3 trillion for the first time.
The rally then spread upstream. $Applied Optoelectronics Inc.(AAOI)$surged roughly 17%, LITE and COHR gained close to 10%, while $闪迪(SNDK)$ rose about 6%.
The logic behind this move is straightforward:
Once the market believes the Mag 7 will not suddenly cut AI capital spending, money returns to cloud companies first—and then spreads into optical networking, storage, and other beaten-down AI infrastructure stocks.
Why did the Mag 7 suddenly surge?
The first reason was lower macro risk.
Trump paused another planned military strike against Iran, sending oil prices down roughly 5% and pulling Treasury yields lower.
The market had been worried that higher oil prices would revive inflation and force the Federal Reserve to keep rates elevated for longer. As that risk temporarily eased, investors moved back into longer-duration technology stocks.
But the durability of this rally will ultimately depend on something more important: Amazon and Microsoft are beginning to prove that AI capital spending can translate into actual revenue.
AWS revenue grew roughly 37%, showing a clear acceleration, while Microsoft gave no indication that it plans to slow AI investment.
The market is starting to accept a different interpretation of AI spending. The billions being poured into data centers are not simply disappearing into a black hole. They are beginning to generate real cloud revenue.
Previously, every CapEx increase was immediately viewed as negative because it meant weaker free cash flow, higher depreciation, and pressure on margins.
Now the market is starting to see the other side:
Cloud growth is accelerating again, and AI investment is beginning to produce measurable returns.
That is the most important reason behind today’s Mag 7 rally.
Why did the rally spread into optical networking?
Optical networking is one of the most direct upstream beneficiaries of AI capital spending.
Building larger GPU clusters requires more than chips. These systems also need faster switches, optical transceivers, and fiber connections. The larger the cluster becomes, the more important internal data transmission becomes—and the greater the role of optical networking inside the data center.
Once Amazon and Microsoft confirmed that AI data-center expansion was continuing, investors naturally began buying back high-beta names such as AAOI, LITE, and COHR.
AAOI jumped roughly 17%, LITE and COHR gained close to 10%, and Corning rose about 6%.
The immediate catalyst was an analyst upgrade for Corning. But without the broader rally in the Mag 7 and the Nasdaq, one upgrade would not have produced such strong sector-wide momentum.
Today’s move was really driven by three forces:
First, optical stocks had already fallen too far. Second, short sellers began covering. Third, the Mag 7 restored confidence that AI capital spending would not suddenly disappear.
I would describe this as the first meaningful recovery in optical networking, but not yet the beginning of a confirmed new uptrend.
The real confirmation still has to come from AAOI, LITE, and COHR through stronger orders, improving margins, and solid customer guidance. The fundamentals must catch up with the share-price rebound before this can be called a true reversal.
Why did storage stocks begin to follow?
An AI data center cannot run on GPUs and optical components alone.
Model weights, training datasets, inference caches, and generated content all need to be stored. As long as the Mag 7 continues expanding data-center capacity, demand for enterprise SSDs and high-capacity HDDs cannot remain disconnected from AI capital spending forever.
SNDK’s roughly 6% gain was another example of money spreading into one of the hardest-hit parts of the AI supply chain.
However, the storage sector is still showing clear internal divergence.
SK Hynix and Samsung continued to fall. The Korean market is still dealing with deleveraging, while CXMT’s expansion has intensified concerns about future supply and competition in traditional DRAM.
Kioxia, however, is trading on a different story.
Kioxia and SNDK are primarily exposed to NAND, while CXMT’s current expansion is mainly a threat to DRAM. Kioxia has also announced a major share buyback and stock split while continuing to confirm that AI data-center demand for NAND remains intact.
That sends two important signals:
NAND profitability is still strong enough to support shareholder returns, and enterprise SSD demand has not suddenly disappeared just because storage stocks collapsed.
Kioxia and SNDK jointly operate NAND manufacturing facilities in Japan, which closely links their production capacity, costs, and technology roadmaps. A stronger outlook from Kioxia therefore improves expectations for SNDK ahead of earnings.
But I do not think Kioxia alone explains SNDK’s rally.
The more important factor was the Mag 7 surge, which restored confidence in the broader AI infrastructure trade. Kioxia repaired expectations for NAND, while the Mag 7 restored risk appetite.
Together, those two forces drove SNDK’s rebound.
The hurdle for SNDK earnings just moved higher
SNDK is scheduled to report earnings after the market closes on August 5.
The company previously guided for revenue of $7.75 billion to $8.25 billion, adjusted EPS of $30 to $33, and a non-GAAP gross margin of 79% to 81%.
The problem is that Wall Street expectations have already moved above the company’s original guidance. Consensus estimates now sit at roughly $8.3 billion in revenue and $34.24 in EPS.
That means simply reaching the high end of management’s guidance may no longer be enough.
If revenue exceeds $8.5 billion, EPS reaches $35 to $36, gross margin remains above 81%, and the next-quarter outlook shows continued growth, I would consider that a genuine beat.
If revenue lands between $8.2 billion and $8.4 billion with EPS between $33 and $35, the result would be closer to expectations. The stock could initially jump and then turn volatile.
If revenue falls below $8.2 billion or next-quarter guidance shows a clear slowdown, the market will quickly return to the idea that NAND pricing has already peaked.
My view is that SNDK has a good chance of beating its original guidance, but repeating last quarter’s extraordinary level of upside will be much harder.
The most important questions are not about the quarter that has already ended. Management needs to prove three things:
Enterprise SSD orders are still growing, gross margins near 80% are not the top of the cycle, and long-term supply agreements can genuinely lock in volume and pricing through 2027.
My view
Today was more than a simple oversold bounce.
The Mag 7 rally showed that investors are beginning to accept AI capital spending again. The surge in optical stocks showed money moving upstream through the data-center supply chain. The gains in SNDK and Kioxia suggested that the market may have become too pessimistic about storage demand.
The capital flow is now becoming clear:
Mag 7 validates continued AI spending → cloud valuations recover → optical networking rebounds first → storage begins to follow.
But today only proved that the AI trade is not over. It did not prove that optical networking and storage have completed a full reversal.
The real test will be whether upcoming earnings can turn this rally into fundamental confirmation.
For SNDK, the Mag 7 surge has created a better environment heading into earnings—but it has also raised the bar.
SNDK no longer needs to prove that AI requires more storage. It needs to prove that it can earn more from this AI investment cycle than the market already expects.
Comments