As Anthropic Races Toward IPO, Tencent, Alibaba, and ByteDance Reportedly Plan to Buy Over 3 Million Nvidia GPUs

[[**Anthropic races toward the largest IPO ever, as Tencent, Alibaba and ByteDance are reported to plan purchases of over 3 million Nvidia GPUs worth more than $35 billion; AI Agents drive up inference costs, while Kimi's quota controversy appears alongside delivery and return risks.**

**Key points**

**1. Agent tasks amplify inference consumption**

After Kimi K3.1 Pro launched, quotas were consumed too quickly; Moonshot AI apologized and compensated 25%; Claude users also complain that multiple sub-agents quickly exhaust quotas.

**2. Big tech's GPU buying spree aims to lock in inference compute**

According to Reuters, demand from ByteDance, Alibaba and Tencent for the RTX PRO 5500 totals more than 3 million units, estimated at over $35 billion; under export restrictions, deployability matters more than top-tier specifications.

**3. The RTX PRO 5500 is not a drop-in replacement for data-center GPUs**

The card has 84GB of memory, about 1398GB/s, and 600W, and mainly handles AI inference and professional workloads; efficiency depends on the model, quantization, concurrency and software adaptation.

**4. Domestic alternatives gain share, but competition still hinges on ecosystem delivery**

IDC says China's AI accelerator shipments in the first half of 2026 were about 2.008 million units, with Huawei accounting for 28%, and shipments of other types of accelerators already exceeding GPUs; constrained Nvidia supply opens a window, but the long term depends on products and delivery.

**5. A $35 billion order does not equal revenue**

If the procurement amount is amortized over five years, equipment costs are about $7 billion per year, plus servers, power and operations; the order must go through delivery, deployment, paid usage and revenue recognition before returns can be verified. 2026-10-28 07:54 With the largest IPO in history approaching, Tencent, Alibaba and ByteDance snap up GPUs: Anthropic and Chinese big tech are together getting hooked on compute Singularity Research Society Quick Overview

This article comes from the WeChat public account: Singularity Research Society, author: Singularity Team, editor: Muren

As Anthropic races toward the largest IPO in history, its prospectus puts compute investment and supply risks front and center; Tencent, Alibaba and ByteDance spend $35 billion to snap up GPUs, and Kimi apologizes to users for quotas being consumed too quickly... As AI Agents begin to take on increasingly complex tasks, inference demand, computing costs and commercial returns are becoming the same exam question.

The more capable AI Agents become, the faster users' quotas may burn.

On October 24, Moonshot AI released Kimi K3.1 Pro. This is the first model in the K3.1 series, with a parameter scale reaching the 3-trillion level and support for a 1-million-token context; the company highlighted the model's capabilities in programming, 3D creation and long-horizon Agent tasks. But after the new model launched, some paying users found that their quotas were nearly used up before their tasks were finished.

On October 26, Moonshot AI issued a public apology, acknowledging that Kimi members using K3.1 Pro experienced rate limiting and "unnecessary quota consumption," and compensated subscribers within their validity period with quota equivalent to 25% of the monthly total.

Where did the problem come from? When a user sends one instruction, multiple sub-agents may already have been launched in the background, separately responsible for breaking down the task, parallel retrieval, analyzing materials, checking results and summarizing output. What the user sees is one task, but what the model actually executes may be a series of calls. The more complex the task, the more tokens and computing resources are consumed in the background.

Similar complaints have also appeared in Claude's user community.

A user posted on Reddit that when Claude Opus 5.5 organized 87 text materials, it automatically launched 12 peer sub-agents and exhausted the entire 5-hour usage quota in 5 minutes; another user reported that using Opus 5.5 for coding tasks for less than 30 minutes triggered the 5-hour quota limit.

These cases may not be universal, but complaints about quotas being consumed too quickly have brought a real problem with Agent products to the fore: how much computing cost do background calls add, and how much actual benefit do they bring users?

Yesterday, Anthropic submitted an IPO application to the U.S. Securities and Exchange Commission. In the prospectus, obtaining sufficient computing resources and continuing to invest in computing infrastructure were included in the company's operating and risk framework, and the company disclosed its arrangements for using various computing resources such as AWS Trainium, Google TPU, Nvidia GPUs and AMD GPUs.

Users care about whether their quotas are sufficient; model companies need to think about how much compute is needed to support these complex tasks, and whether they can obtain sufficient commercial returns from them.

A potential procurement order recently disclosed by Reuters pushed this supply-demand pressure to a larger industrial scale.

According to Reuters, citing people familiar with the matter, demand from ByteDance, Alibaba and Tencent for Nvidia RTX PRO 5500 totals more than 3 million units. Based on a list price of about $11,400 each, the potential order size exceeds $35 billion. At the same time, Nvidia cannot guarantee it will meet all demand, and actual delivery volumes may be far below order intentions.

The potential $35 billion order and users' complaints that quotas burn too fast appear to be two different problems, but both are related to growing inference demand. As AI moves from answering questions to executing complex tasks, consumption of computing resources increases accordingly, and the pressure extends from users to model companies and the chip supply chain.

To what extent has demand for inference compute actually grown? Why are major Chinese internet companies still collectively purchasing an Nvidia graphics card whose performance is inferior to top-tier data center chips? And how will this scramble for cards affect the commercialization of domestic chips?

Even workstation cards are being snapped up

What are the major companies anxious about?

If we rewind one year, the product route represented by the RTX PRO 5500 might not have attracted such attention.

According to previous Reuters reports, when Nvidia launched restricted products for the Chinese market such as the RTX 6000D, some large technology companies showed little interest, believing the price and performance of such chips did not match, and they also hoped the U.S. government would allow more powerful products to re-enter the Chinese market.

That waiting did not bring a definite result. U.S. chip export restrictions on China persist; exports of products such as the H200 must meet specific licensing conditions, while the more powerful B300 cannot be sold to China under current restrictions. At the same time, AI models and applications are still expanding, companies need to continuously deploy new models and expand service scale, yet compute supply may not keep pace with business rhythms.

In such a supply-demand environment, compute that can be bought and deployed is more important than ideal chip specifications.

What exactly can the RTX PRO 5500 do? In terms of product positioning, the RTX PRO 5500 is a professional workstation GPU based on Nvidia's Blackwell architecture: 84GB GDDR7 memory, memory bandwidth of about 1398GB/s, and maximum power consumption of 600W. Nvidia positions it for professional workloads such as Agentic AI, physical AI, simulation and graphics processing.

Its 84GB of memory and corresponding compute give it the conditions to handle some AI inference tasks. Compared with consumer-grade graphics cards, professional workstation GPUs also have designs and support capabilities for professional scenarios.

But do not confuse it with the H200 or B300. The latter two are aimed at large-scale data center clusters, serving high-intensity AI training and inference, and rely on large-scale interconnection to handle heavier model workloads. The RTX PRO 5500 is closer to a workstation GPU capable of professional computing and some AI inference.

Whether a model can run efficiently depends on model scale, quantization method, parallel deployment, software adaptation and actual throughput.

It can address part of the computing demand, but it cannot be promoted as a replacement for high-end data center GPUs based solely on 84GB of memory. Reuters, citing people familiar with the matter, said Chinese companies are purchasing these graphics cards mainly for AI inference.

Inference occurs after model training is complete: a user question, a code generation, an enterprise knowledge base query, or even an Agent continuously calling multiple models while executing a task all generate inference computing demand.

Training is like concentrated construction, while inference is like daily water supply. Model training usually concentrates a large amount of compute to complete parameter updates in a relatively concentrated period; inference follows product usage, and the more users there are, the more frequent the calls and the more complex the tasks, the more continuously model services must supply compute.

As Agents move from "answering questions" to "executing tasks," inference volume will be amplified layer by layer. A single user request may be split into several stages: retrieving materials, breaking down tasks, calling tools, generating results and checking results. The question companies care about has also changed from "are there enough cards to train models" to "can models be run continuously, stably and economically."

This explains why a product whose performance is inferior to top-tier data center chips is still being snapped up by major companies.

What exactly are these cards used for? Regarding purchase quantities, deployment scenarios and alternatives, we sought confirmation from Tencent and ByteDance. Tencent said it was not clear, while ByteDance and Alibaba had not responded as of publication.

Existing information is not enough to break this order down into the specific business plans of the three companies: it cannot be concluded that ByteDance is necessarily buying for Doubao, or Alibaba necessarily for Qwen. How much each has ordered, where they will be deployed, and whether they have been confirmed for use alongside domestic chips all remain to be confirmed.

The industrial application direction can be confirmed. In theory, such graphics cards can handle online services for small and medium-scale models, some enterprise internal AI applications, Agent task execution, and professional computing workloads with certain memory requirements. For specific tasks, parameter scale, quantization schemes, concurrent requests and latency requirements still need to be evaluated. Some businesses may also adopt multi-card deployment, but multi-card efficiency and cost cannot be directly inferred from single-card specifications.

Inference also does not mean training is completely ruled out. Workstation GPUs can be used for development, testing and some lightweight fine-tuning; but in this purchase, Reuters points to AI inference as the main use. Without more evidence, it should not be understood as major companies sweeping up top-tier pre-training compute.

As for why they have not fully switched to domestic chips, the answer must be viewed in terms of software ecosystems, existing clusters, migration costs and deployment efficiency. Buying Nvidia does not mean the three companies are not using domestic chips; it may simply be a supplement to the existing compute system. What is more worth asking is how much companies are willing to pay for obtainable inference compute, and how large they judge future demand will grow.

We should also make clear that a $35 billion order size does not mean Nvidia has already secured a certain revenue of the same amount.

According to Reuters, related demand from Chinese customers exceeds 6 million units; Nvidia expects global RTX PRO 5500 production in 2027 may not exceed 3 million units. The company also has to balance different product lines and allocate more capacity to the next-generation Rubin platform.

Even if all procurement intentions are fulfilled, delivery will be constrained by production capacity. For buyers, locking in supply in advance is about securing future resources. Even if the final delivered quantity differs greatly from the initial demand, as long as they obtain a portion, it may still supplement compute power, support expansion, or adjust the load on existing clusters.

How much locking in supply is worth also depends on contract terms, the final transaction price, the specific delivery quantity, and whether the products can be deployed as planned. For the capital market, these are the pieces of information that must continue to be tracked to judge the value of the order.

For Nvidia, this order shows that even if some high-end products are subject to export restrictions, the Chinese market still needs its product ecosystem and computing power. As for how much of the demand can be converted into actual revenue, that will also depend on subsequent deliveries.

Buying chips is only the beginning

A $35 billion procurement intention is enough to reflect big companies' expectations for future compute demand, but it cannot directly prove that AI inference in China is already comprehensively oversupplied with demand.

IDC data shows that in the first half of 2026, China's AI accelerator shipments were about 2.008 million units. Among them, Nvidia shipped about 678,000 units, accounting for 34% of the market; Huawei shipped about 572,000 units, accounting for 28%. Alibaba's T-Head accounted for 7%, Cambricon 5%, Baidu Kunlunxin 4%, and Hygon 3%.

During the same period, China's GPU shipments fell 26% year on year to about 991,000 units, while shipments of other types of AI accelerators grew nearly 80% to about 1.016 million units, already surpassing GPUs. IDC expects China's overall AI accelerator shipments to grow 19% in 2026, while GPU shipments may decline 9%.

Therefore, this wave of chip stockpiling cannot simply be understood as a shortage of all types of compute. What enterprises truly need is computing resources that can run their target models, meet concurrency and latency requirements, and be deployed in a timely manner.

A chip's memory, bandwidth, computing power, and software support affect whether a model can run and how efficiently it runs; once it enters actual business operations, servers, networking, scheduling, and operations and maintenance must also be considered. Even if two chips have similar theoretical computing power, actual throughput, response latency, and total cost can differ greatly.

For enterprises that need to launch model services quickly, procurement timing is equally important. After chips arrive, model adaptation, system deployment, and business testing still require investment. If migrating existing systems requires extensive engineering changes, both time and cost will affect the final choice.

This is also one reason Nvidia products remain attractive. Nvidia's long-accumulated CUDA ecosystem, operator libraries, and development tools bring engineering convenience to its products. Enterprises can continue to use some existing development and operations experience, reducing uncertainty in the migration process.

However, even if the chips arrive smoothly, enterprises still need to consider whether the added compute can generate sufficient business revenue.

Assuming the entire batch of chips is ultimately purchased for $35 billion, and only the equipment procurement amount is amortized evenly over five years, that corresponds to about $7 billion in equipment cost per year. This calculation only illustrates the scale of the investment and does not represent the three companies' actual procurement contracts, payment schedules, or accounting treatment.

Moreover, chips are only part of AI infrastructure costs. Servers, network equipment, data centers, power, cooling, and operations and maintenance all generate corresponding expenses. After the equipment is in hand, if utilization is insufficient, the unit computing cost may be high; if model service fees are low and user payment rates are low, added compute may not necessarily generate sufficient revenue.

Therefore, judging whether this procurement is worthwhile requires looking at several operating indicators: the comprehensive cost per million tokens, effective throughput per card, compute utilization, paid invocation volume of AI services, and whether related business revenue can cover depreciation and operations and maintenance expenses.

Different enterprises also have different return paths. Cloud providers can directly generate revenue by selling compute power and model services; platform companies with consumer AI products may generate returns indirectly by improving user retention and driving advertising or transaction conversion. The commercial value of the latter may not be directly reflected in AI product revenue, but it likewise needs to be verified through actual operating data.

Only when the chain from procurement intention to chip delivery, and then to business revenue and profit, is proven can the $35 billion have clearer commercial significance.

Under the chip stockpiling wave

How can the compute dividend of domestic substitution be realized?

This order occurred during a period of continuous adjustment in U.S. chip export policy toward China. Sales restrictions differ across products, and only by understanding these rules can one understand why big companies are stockpiling chips and what competitive environment domestic chips face.

(Note: As of October 2026, specific sales and deliveries remain subject to relevant licenses, supply arrangements, and import policies.)

Nvidia disclosed in its 2026 regulatory filings that although H200-related licenses were approved, the Chinese government's import restrictions still affected actual sales of some products, and the company was unable to sell all licensed products.

This means whether chips can enter the Chinese market depends not only on whether the United States allows exports, but also on buyer qualifications, supply arrangements, and China's import policies. For purchasing enterprises, complying with export restrictions is only the starting point; whether products can arrive on time and be put into business use is equally important.

If Nvidia can stably provide more compliant products in the future, some enterprises may readjust their procurement plans. For customers that have already deeply used Nvidia's software ecosystem, continuing to purchase compatible products can reduce migration and adaptation costs; domestic chip manufacturers, meanwhile, will face more direct competition.

But the resumption of imported supply does not mean that domestic substitution will lose its room. According to IDC data cited by Reuters, in the first half of 2026, Huawei's shipment share in China's AI accelerator market reached 28%; domestic suppliers such as T-Head, Cambricon, Kunlunxin, and Hygon have also occupied a certain share. Domestic manufacturers are expanding shipments and using actual deployments to accumulate software adaptation and cluster operation experience.

For major companies like Tencent, Alibaba, and ByteDance, the stability of computing power supply is also becoming increasingly important. Training, inference, and cloud services all require continuous investment in computing resources, and the policy or production capacity risks faced by a single supply channel may affect the pace of business expansion. Therefore, enterprises are motivated to evaluate different chip solutions and allocate computing power according to task requirements.

Restricted Nvidia supply has opened a market window for domestic chips, and whenever news of large computing power procurement is released, the capital market looks for potential beneficiary directions, but from procurement intention to listed company performance, there are still multiple links that need to be verified.

Server and system integrators may benefit from subsequent deployment, but the RTX PRO 5500 is a workstation GPU, and its deployment method differs from large data center GPU clusters. Based on procurement quantity alone, it is impossible to infer server demand, let alone directly determine the corresponding rack, network, and cooling solutions.

Power supply, cooling, and data center infrastructure companies may also gain incremental demand, but the specific scale depends on chip deployment methods, power configurations, and whether existing facilities can accommodate them. For high-speed interconnection links such as optical modules, judgment needs to be more cautious. Large AI clusters may drive demand for high-speed networks, but this round of procurement does not equal data center interconnection orders of the same scale. Whether related companies benefit still depends on whether specific projects and equipment procurement can be implemented.

Domestic chips and their supporting ecosystems also have opportunities to undertake part of the new demand, but from customer testing to large-scale deployment and then to revenue recognition by listed companies, time is still needed. To judge the degree to which specific enterprises benefit, their product capabilities, customer orders, mass production delivery, and financial contribution should be further verified.

Procurement demand, chip delivery, infrastructure construction, listed company orders, and final revenue are five different links. The computing power concept being related to this order does not mean that the relevant companies will necessarily benefit; only by finding specific business connections can the investment logic be supported.

Ending

The $35 billion procurement intention reflects leading enterprises' expectations for future inference demand and also exposes real problems such as restricted imported chip supply and domestic solutions still requiring continuous verification.

For purchasing enterprises, securing computing power resources is a prerequisite for business expansion. Subsequently, they still need to complete deployment, control inference costs, and make the new computing power serve actual business.

For domestic chip manufacturers, the supply gap provides opportunities, but long-term competition still depends on products, software ecosystems, and delivery capabilities. For the capital market, order scale cannot replace judgment about actual projects and performance.

Has published 22 articles on Huxiu. This content is authorized by the author for publication, and the views represent only the author and do not represent Huxiu's position. If you have objections or complaints about this manuscript, please contact tougao@huxiu.com. Everyone who is changing and wants to change the world is on the Huxiu APP[ ]]]