Where did Tencent and Alibaba's Capex go?

After reviewing the financial reports of Tencent and Alibaba, both have relatively solid fundamentals. If we must compare, Tencent is better than Alibaba. Especially Tencent, not only are its main businesses in gaming and advertising solid, but its investments and capital planning in AI must be reiterated: Tencent is still the same Tencent, stable as ever. Alibaba, on the other hand, has focused heavily on the growth of its cloud business, which has high expectations. Before it can "gradually accelerate in the coming quarters," it is quite normal to have some digestion and adjustment.

However, the biggest mystery this quarter is that both companies' capex has decreased significantly quarter-on-quarter. Where did the money go? After listening to Alibaba tonight, the conclusion is the same: there is money that cannot be spent.

The $5.5 billion provision by NV, we have calculated, translates to an impact of about $15 billion in revenue when converted to H20 terminal prices, which is about 100 billion RMB. The provision by NV is essentially the expenses saved by BAT.

If 40% of that belongs to Tencent and Alibaba, approximately 40 billion occurred between January 31 and April 30 (the quarter of NV's financial report). If that 40 billion were to be added back to this quarter's capital expenditure, it would likely be more than enough.

Unfortunately, with SMIC's production capacity fully ramped up, it is indeed impossible to squeeze out 40 billion in domestic alternatives within three months...

What to do next? Martin has already provided the answer: "Start with software." If hardware is insufficient, optimize to compensate, and continue to promote the cost reduction and efficiency improvement characteristics of Chinese AI. From Martin's tone, Tencent is indeed relatively rational and cautious in its spending. If AI remains as it is, that would be fine, but clearly, the war for Agents has just begun, and we will see a dazzling competition for Agents in the second half of the year, with demand only increasing, and possibly accelerating.

So, H is not allowed to be used, and H20 is not allowed to be sold. What to do? The demand is right there; not only do other domestic cards besides the "Juju" want a piece, but Jensen Huang also wants in. Therefore, reflecting on the BIS "long-arm document" from a couple of days ago, it is both a demand and a demand. It wants to suppress "Juju," curb Chinese AI, and also wants to profit from you.

This point is worth clarifying: even if the U.S. strategy is to achieve AI monopoly by selling the American technology stack globally, the ultimate goal is to harvest the global market through selling terminal applications. I can sell you a hoe, but you absolutely cannot have the ability to produce your own hoe and compete with me for a living.

It seems that things are stuck here. Is there no way out? Perhaps not. We will not elaborate on domestic alternatives, southern factories expanding production, ASIC breakthroughs, and so on. Tonight, there is one thing that not many people are paying attention to. DeepSeek published a paper, also signed by Liang Wenfeng, mainly sharing the importance of hardware-software synergy discovered in V3. At the end of the paper, regarding the current chip hardware bottleneck, it pointed out the directions for breakthrough innovations that DeepSeek believes can be achieved.

Relying solely on BAT's spending and southern factories expanding production is just a linear catch-up approach. After all, if we continue along the American path, relying on miracles, NV system advantages, coupled with embargoes, we may always be led by the nose. The synergy of models + chips + systems, optimizing and innovating together, and finding a path suitable for China's own ecosystem may be the true breakthrough to overcome this systemic blockade Information Equality, Original Title: "Where Did Capex Go?"

Risk Warning and Disclaimer

The market carries risks, and investment should be cautious. This article does not constitute personal investment advice and does not take into account the specific investment objectives, financial conditions, or needs of individual users. Users should consider whether any opinions, views, or conclusions in this article are suitable for their specific circumstances. Investment based on this is at one's own risk