In mid-May, the U.S. Department of Commerce approved procurement by about 10 Chinese companies for NVIDIA H200 artificial intelligence chips. Lenovo Group was identified as one of the distributors authorized to sell H200 in China. After the news was released, Lenovo’s stock price continued to rise steadily over just a few weeks. As of May 29, 2026, after the market opened, Lenovo’s gain at one point exceeded 30%, with a high of 25.7 HKD, before narrowing to 20%; it was temporarily quoted at 23.76 HKD. Since late April, Lenovo’s stock price has doubled.
This round of上涨行情 is not driven by a single piece of news, but rather by a structural re-rating formed by the combined impact of multiple factors, including H200 distribution authorization, performance exceeding expectations, and the restructuring of the AI compute supply chain.
What does the implementation of H200 distribution authorization reflect about the shift in export control policy?
Judging from changes in the regulatory framework, in January 2026 the U.S. Department of Commerce’s Bureau of Industry and Security (BIS) issued new rules, changing the licensing review policy for NVIDIA H200 and similar chips from “presumed denial” to “case-by-case review.” This policy adjustment means that although the overall direction of export controls has not changed, the U.S. has opened a limited policy window for sales of certain compliant chips to China.
On May 14, 2026, during Trump’s visit to China, the H200 sales license list was officially finalized. Approved customers include internet companies such as Alibaba, Tencent, ByteDance, and JD.com. Lenovo and Foxconn received distribution authorization. Under the license terms, each approved customer may purchase up to 75,000 H200 chips—either directly from NVIDIA or via authorized distributors.
However, the so-called “loosening” of policy falls far short of unconditional clearance. Reports indicate that although the U.S. has issued licenses, as of mid-May no H200 chips had been actually delivered to Chinese customers. The U.S. also set stringent additional conditions, including requiring NVIDIA to remit 25% of relevant sales revenue to the U.S. government and requiring Chinese buyers to disclose customer information. This means that actual H200 sales still face a dual barrier: approvals from the Chinese side and commercial decision-making.
What role does Lenovo play in the AI compute supply chain? Why has it become the core beneficiary of this rally?
Lenovo’s identity on the approved list is especially significant. It is not only an approved procurement party; it has also been officially confirmed as a distributor. This dual role means Lenovo holds two layers of value at the same time: as a server manufacturer, it can directly integrate H200 chips to launch high-performance AI servers and meet large-model training needs; as a distribution channel, it can leverage its supply-chain network to distribute compliant compute to more government and enterprise customers.
In a complex export control environment, this “compliance bridge” hub position is scarce. Lenovo’s AI infrastructure buildout is not just a matter of headlines. The full-year performance report for fiscal year 2025/26 released on May 22 shows the group’s full-year revenue reached RMB 589.9 billion, up 20.3% year over year; adjusted net profit grew 42.1% year over year; both hit historical highs. In the fourth fiscal quarter, revenue grew 27% year over year on a quarter-by-quarter basis, AI-related business revenue grew 84% year over year, and its share of total group revenue increased to 38%. AI-related revenue doubled on a full-year basis year over year.
Meanwhile, orders for AI server inventory reserves totaled RMB 140.0 billion, up 50% year over year. These figures indicate that Lenovo’s distributor identity is not a catalyst created out of thin air—it is built on fundamentals of rapid growth in its existing AI infrastructure business.
Beyond H200 distribution rights, what other factors support Lenovo’s doubled rally within one month?
Before the H200 distribution rights news, Lenovo’s stock had already entered an upward channel. On March 17, at NVIDIA’s GTC conference, Lenovo and NVIDIA jointly released the Hybrid AI solution, laying a technical foundation for deep cooperation between both parties in AI infrastructure. Strong financial results were another key driver: fourth-quarter revenue was $21.6 billion, above market expectations of $18.7 billion; Non-HKFRS net profit was $559 million, up 101% year over year, about 75% higher than Bloomberg’s consensus expectation. After that, multiple institutions collectively raised their target prices. Goldman Sachs raised its target price sharply to 27 HKD and China International Capital Corporation raised it to 20 HKD.
In addition, Lenovo’s AI strategy is extending from servers to a broader industrial chain. On May 27, Lenovo released what it said is the world’s first commercial AI edge device, the “Bainying AI Host” (百应 AI 主机), rolling out a Token Plan compute service model to standardize compute and deliver it on a subscription basis. On May 28, Lenovo announced it would invest in Tianjin to build a new next-generation AI compute product R&D and manufacturing center. The new production lines are planned to go into mass production in 2027, and its general server production lines are also planned to go into mass production in Tianjin in September 2026. These long-term deployments further strengthen market expectations for sustainable growth in its AI business.
How does the global AI compute shortage lead to value re-assessments across markets? Is Lenovo’s rise an isolated case or a snapshot of an industry-wide trend?
Lenovo’s rally is not an isolated case; it is a microcosm of a global structural shortage of AI compute capacity. From late 2025 to early 2026, as AI large models shift from training to large-scale inference deployment, global compute demand entered an exponential growth channel. IDC forecasts that the number of active AI agents worldwide will grow from 28.6 million to 2.216 billion by 2030—nearly 80 times. China’s National Data Bureau shows that as of March 2026, China’s average daily Token calls exceeded 140 trillion, up 1,400 times compared with early 2024.
While demand grows explosively, supply faces a structural stalemate. In 2026, the global market supply-demand gaps for DRAM, NAND, and HBM are projected to be 4.9%, 4.2%, and 5.1%, respectively—each the highest level since 2011. The HBM capacity gap is even higher at 50% to 60%. The three major memory foundries will direct 70% of incremental capacity to the HBM segment. SK hynix’s HBM capacity before the end of 2026 has already been locked in by major customers early. The spot GPU market is also in short supply: the rental price for one-year H100 contracts has risen by nearly 40% within half a year.
This supply-demand imbalance has appeared across multiple markets: Samsung Electronics and SK hynix’s market caps both surpassed $1 trillion in sequence, and Micron Technology has also joined the trillion-dollar club. Lenovo’s rise follows the same logic line extension: as a core supplier of AI compute infrastructure, its value has been re-priced in the macro narrative of compute scarcity.
Is there any capital or narrative transmission between the AI hardware market and the crypto market? What does a compute shortage mean for the crypto market?
Notably, the narrative of compute shortage has spread from traditional markets into the cryptocurrency asset space. According to Gate market data, as of May 28, 2026, decentralized compute token RENDER was priced at $2.1023, with a gain of 21.68% over the past 30 days; decentralized storage token FIL was priced at $1.0414, with a gain of 12.05% over the same period. Meanwhile, AI sector tokens rose collectively in late May: WLD surged 32.3% within 24 hours, NEAR rose 81% over one week, and FET rose 17.44%.
This resonance reflects two different transmission logics. On one hand, a shortage of AI compute increases market attention to AI infrastructure projects, and decentralized compute networks are viewed as one of the complementary solutions to ease centralized compute bottlenecks—prompting the market to re-assess the potential value of such projects. On the other hand, the crypto market has a naturally high sensitivity to macro narratives; strong performance in the AI hardware sector easily draws capital attention and trading activity toward the “AI + Crypto” theme. The two transmission pathways reinforce each other, forming theme rotation across markets.
From distribution authorization to actual sales, what uncertainties still face Lenovo’s AI compute story?
Multiple uncertainties coexist. First, although H200 export licenses have been approved by the U.S., the stance of China’s approval process remains unclear. Analysts suggest China’s authorities may want to leave more market space for domestic substitution; Huawei Ascend 910C has already approached H200 in technical performance. It is expected that China’s domestic high-end AI accelerator shipments in 2026 will reach 21.23 million units, and Huawei’s share in the local market has already exceeded 50%. This could limit the actual purchasing intent of Chinese customers even if the U.S. approves.
Second, the international geopolitical environment continues to change dynamically. In April 2026, multiple U.S. lawmakers proposed the MATCH bill draft in an attempt to further tighten export controls on semiconductor equipment to China. While H200 currently enjoys “limited exemption” under the existing policy framework, the direction of export control policy evolution remains highly uncertain, putting the long-term sustainability of H200 distribution to the test.
Additionally, market expectations for growth in AI compute have already been significantly reflected in current valuations. Whether Lenovo’s AI server reserve orders can be converted into revenue and profit as scheduled, and whether distribution channels can remain accessible amid changes in the regulatory environment, are still core variables the market needs to keep tracking.
FAQ
Q: Has Lenovo started selling H200 chips to Chinese customers?
As of mid-May 2026, despite the U.S. having issued export licenses, H200 chips have not yet completed actual delivery to the Chinese market. China’s approval stance and the U.S.’s stringent additional conditions (including the 25% revenue share and customer information disclosure requirements, etc.) remain major obstacles to actual sales.
Q: Is Lenovo’s stock rally driven only by the H200 distribution rights news?
H200 distribution rights are a direct catalyst, but the core drivers of the stock rise also include: fiscal year 2025/26 performance far exceeding expectations (revenue up 27% year over year, net profit up 101%), AI server reserve orders totaling RMB 140.0 billion (up 50% year over year), multiple institutions collectively raising target prices, and the implementation of long-term capacity deployments such as the Tianjin AI compute R&D and manufacturing center.
Q: How might the future direction of export control policy affect Lenovo’s AI business?
H200 export licensing is currently under a “case-by-case review” policy framework, which could still be adjusted as U.S.-China relations change. If the MATCH bill proposed in April 2026 passes, it would further tighten exports of semiconductor equipment to China. Lenovo’s AI server business relies heavily on NVIDIA chips; the sustainability of distribution authorization is the core risk variable. Progress on domestic substitution will directly affect its long-term bargaining power and market space.
Q: What are the specific transmission paths of a compute shortage to crypto assets?
A compute shortage transmits to crypto assets mainly through two paths: first, increased attention to decentralized compute networks (such as Render Network) and decentralized storage (such as Filecoin) as complementary solutions boosts re-assessments of relevant token valuations; second, strong performance in the AI hardware sector triggers rotation of capital attention toward the “AI + Crypto” theme, and the crypto market—being a higher risk-appetite sector—reflects this theme first.
