Why Alibaba Stock Is Rallying While US Chip Stocks Break

Alibaba trades around $120.26 and is up roughly 17% month to date, which puts it about 10% above the $109 level it jumped to on July 8 during its best single session in ten months. On Friday, July 17, 2026, the same week that produced that move handed the stock another one, and it came from an unusual place. Chinese AI lab Moonshot AI, which counts Alibaba among its backers, released Kimi K3, a 2.8 trillion parameter model and the largest open-weight release anyone has shipped.

US semiconductor names went the other direction on the news. Taiwan Semiconductor fell 7% on Friday even after reporting a 77% jump in quarterly operating profit, SoftBank fell 9.0%, and the PHLX Semiconductor Index entered a bear market. Alibaba rallied through all of it.

- BABA: around $120.26, up roughly 17% month to date

- NVDA: around $203.48

- AMD: around $503.10

- INTC: around $95.70

- GOOGL: around $348.78

- Crypto context: BTC near $64,785 and flat, ETH near $1,878

That split is the whole story, and it is not a rounding artifact. Here is what the market is actually pricing, and why it changes how you should think about every AI position you hold.

What Kimi K3 Actually Changed

Open-weight means the model weights are published and anyone can download, run, fine-tune, and deploy them without paying the lab that trained them. At 2.8 trillion parameters, Kimi K3 is the biggest model ever released under those terms, and Moonshot AI put it out through its public model repositories rather than behind a metered API.

The reason a chip index breaks on that news takes one step of reasoning. The US AI trade is built on the assumption that frontier capability stays scarce and expensive, because capability is a function of compute, and compute is sold by a small number of firms with pricing power. TSMC posting a 77% operating profit jump and falling 7% anyway is the cleanest evidence you will get that the market was not repricing this quarter's earnings. It was repricing the durability of the moat behind them.

And the Chinese AI trade is built on the opposite bet. Give the models away, win the developers, win the cloud workloads, win the applications sitting on top. If a capable frontier model costs nothing to license, the economic value stops accruing to whoever sold the training silicon and starts accruing to whoever owns distribution.

Why Alibaba Specifically Sits on the Right Side of That Trade

Alibaba is not a chip company and it is not a pure AI lab, which is exactly why it caught the bid. It owns cloud infrastructure to serve models from, a commerce business with hundreds of millions of users to deploy them into, and a payments network underneath both. Those are distribution assets, and distribution is what open weights make valuable.

It also holds a stake in the lab that just shipped the most discussed model in the world, which matters for narrative more than for the income statement right now. Backing a lab is not owning one, and that distinction is worth holding onto.

Moonshot AI has signaled plans for a Hong Kong IPO within six months, which would give public investors a way to price the lab directly instead of expressing the view through Alibaba. That is a double-edged catalyst. A successful listing validates the asset, and it also hands holders a cleaner instrument, which can pull the narrative premium back out of BABA.

The Bear Case on Alibaba Deserves Equal Weight

Here is where most coverage of this move gets lazy. The bull thesis is genuinely interesting, and the pushback against it is also real.

Brokers were cutting price targets on BABA in early July, before the Kimi K3 move, and the reason they gave was AI capex drag on near-term earnings. Morgan Stanley, Citi, Daiwa and HSBC all trimmed targets into the $170 to $192 range while keeping Buy ratings. The Street still likes the asset and thinks the platform bet works, and it also thinks the spending required to get there compresses the next several quarters of profitability. Building cloud capacity to serve free models is expensive, and that cost lands on margins long before the platform revenue shows up.

Political and regulatory risk is the other half, because Chinese technology equities carry a category of exposure that US names do not and it never shows up in a P/E ratio. Alibaba's $600 million DOJ settlement on July 8 resolved historical illegal-pharma listings on its platforms, and a separate Pentagon-list matter sits under a court stay. A stay is a pause, not a resolution. Both items are background rather than this week's news, and together they describe the standing risk profile.

Then there is the simplest problem. Open-weight models are hard to monetize directly, because the thing you gave away is the thing everyone wanted, and a stock that has added 17% in a month on a narrative can return that 17% on the same narrative cooling.

How the AI Complex Splits From Here

The useful framing is a stack. Value has to land somewhere in it, and Friday was the market voting on which layer.

The table is a map, not a forecast. Nobody has proven that free frontier models actually kill silicon demand, and the counterargument is decent, because cheaper models can drive far more inference volume and inference runs on chips too. That is the bull case for the semis surviving this, and it is why buying the chip selloff is a defensible trade rather than an obviously wrong one.

What is no longer defensible is treating "AI" as one directional exposure. The NVIDIA stock outlook for 2026 and the Alibaba thesis are now, at least partially, opposite trades, and the same tension runs through Micron's memory position, the Samsung and Broadcom semiconductor picture, and the Marvell AI outlook. All of them sit on the layer that just got questioned. You can follow the damage at the index level through the SOX index constituent data and track BABA's own reaction on its Yahoo Finance quote page.

What This Means If You Trade Crypto

Bitcoin sat near $64,785 and flat through Friday's equity move, and ETH near $1,878 did the same. Crypto did not participate in the repricing, which is the first thing worth noticing. This was an equity-complex event about where AI margins land, and it did not read as a broad risk-off signal.

The read-through for traders is structural rather than directional. If you have been expressing an AI view through crypto proxies, GPU-adjacent tokens, decentralized compute networks, or AI agent tokens, most of those sit on the compute-scarcity side of the stack, and Friday was a vote against compute scarcity. That does not invalidate them, and it does mean the correlation you assumed between "AI adoption grows" and "my AI token goes up" now has a regional and structural filter on top of it.

BABA is the clean way to take the other side. It is a tokenized-stock perpetual on Phemex, so the non-US leg of the AI trade is expressible in the same account and the same collateral as your Bitcoin position, without a separate equities broker or a Hong Kong listing account.

The Bottom Line

Watch the gap between BABA and the SOX index rather than either one alone. If Alibaba holds above the $109 July 8 breakout level while the semiconductor index stays in bear-market territory, the transfer-of-value thesis is intact and the market is still voting for the platform layer. If BABA loses $109 while chips recover, Friday was a one-day narrative trade and the capex drag the brokers flagged is what actually matters.

The Moonshot AI Hong Kong listing inside six months is the next real catalyst, and it is the one that can go either way. A strong debut validates the asset and hands investors a cleaner instrument than owning Alibaba as a proxy. Watch the broker targets too. Those $170 to $192 numbers were set before Kimi K3, and the first bank to revise on the platform thesis rather than the capex drag will tell you the Street has changed its mind about which layer wins.

Frequently Asked Questions

Why is Alibaba stock going up in July 2026?

The July 8 jump of about 11% to $109 came alongside the $600 million DOJ settlement clearing a legal overhang, and the move since then reflects Alibaba's position in the AI stack after Moonshot AI released Kimi K3 on July 17. The market is treating Alibaba's cloud and distribution assets as the beneficiary of cheap open-weight models rather than a victim of them.

Does Alibaba own Moonshot AI?

Alibaba is one of Moonshot AI's backers and holds a stake, which is materially different from ownership or control. The planned Hong Kong IPO within six months would let investors price Moonshot directly instead of through Alibaba, which is worth watching if you are holding BABA specifically for the AI exposure.

Why did TSMC stock fall despite strong earnings?

Taiwan Semiconductor reported a 77% jump in quarterly operating profit and fell 7% anyway on July 17, which tells you the selling was about future pricing power rather than the quarter that just closed. If frontier AI capability becomes free and open-weight, the scarcity premium built into semiconductor valuations gets questioned regardless of current results.

Is Alibaba stock a buy at $120?

The honest answer depends on which risk you are willing to hold. Broker targets from Morgan Stanley, Citi, Daiwa and HSBC sit in the $170 to $192 range with Buy ratings attached, and the same firms cut those targets in early July on AI capex drag, so the Street sees upside and near-term margin pressure at the same time. The Chinese-equity political and regulatory risk is separate from all of that and does not go away with a good quarter.

This article is for informational purposes only and does not constitute financial or investment advice. Cryptocurrency trading involves substantial risk. Always conduct your own research before making trading decisions.