Economic Observer Follow
2026-11-20 13:09

Wen/Zhu Zhaoyi
On November 12th local time, US chip giant Nvidia (NVDA) announced that its co-founder Huang Renxun has resigned from his positions as President, CEO, and board member, effective immediately.
Subsequently, Washington successively adjusted the rules of AI (artificial intelligence) cooperation. On November 15th, the United States signed the San Francisco Declaration with countries such as the United Arab Emirates, Qatar, and Israel, incorporating sensitive technologies and their dependent energy, communication, logistics, and other infrastructure into security cooperation. On the 16th, the US Department of Commerce's Bureau of Industry and Security announced remote computing control rules, further restricting specific entities from accessing advanced computing power through overseas data centers.
In the Middle East, Huang Renxun's departure affects two different interests. Israel cares about the research and development team and the planned 10000 person campus; Gulf countries are concerned about whether chips can be delivered on schedule, which customers the data center can serve, and how much return can be obtained from investing in the US AI industry.
They are all closely related to Nvidia, but stand at different positions in the industry chain. The change of leadership of NVIDIA and the adjustment of rules in the United States are happening simultaneously, and the nerves in the Middle East are also tense.
1、 NVIDIA's Second Home
Huang Renxun has repeatedly referred to Israel as Nvidia's' second home '. This closeness lies in the significant weight that the Israeli subsidiary holds in Nvidia's global business.
In 2019, Nvidia announced the acquisition of Israeli chip company Mellanox for approximately $6.9 billion, and the transaction was completed in 2020. Mellanox founder Eyal Waldman and Huang Renxun have been working together for many years, and their technological complementarity and personal trust ultimately led to the merger.
This transaction complements a key capability of the AI cluster. Training models with tens of thousands of GPUs requires continuous data exchange. Communication cannot keep up, and even the most expensive processors have to wait. Mellanox's InfiniBand and high-speed Ethernet technology enable more computing nodes to collaborate effectively.
Nvidia is therefore better equipped to sell chips, networks, and software as a complete system. Customers need to compare both the price of a single chip and the utilization rate, training time, and renovation cost of the entire system. Even if competitors offer cheap GPUs, it may not be easy to replace the entire deployment.
The original team has also entered the management core. Michael Kagan, co-founder of Mellanox, serves as NVIDIA's Chief Technology Officer, while Gilad Shaina serves as Senior Vice President of Network Business. Nvidia disclosed in August this year that its Israel business has approximately 6200 employees, covering functions such as research and development, operations, and sales.
The Kiriyation campus located in northern Israel represents the next phase of Nvidia's expansion plan for its Israel business. The company's announced plan involves a building of approximately 160000 square meters that can accommodate around 10000 employees. It was originally scheduled to start construction in 2027 and be put into use in 2031. The Israeli land management department has provided land incentives, hoping to drive employment, taxation, and surrounding commercial development in the north.
Another noteworthy project is in Mevo Carmel. According to Israeli financial media reports, the estimated total investment in local server facilities and computing equipment is about 1.5 billion US dollars, planned to accommodate Blackwell processors and next-generation supercomputers, support Nvidia's internal research and development and chip testing, rather than renting computing power to external customers.
These uses are very important. The Israeli team is involved in the development and testing of the next generation of products, and the technical personnel are constantly accumulating experience in this process. The mutual strengthening of facility investment and research and development capabilities makes it more difficult for local teams to be quickly replaced by other locations.
After Huang Renxun left, the most important thing to confirm was the unfinished expansion budget. Israeli financial media Calcalist analyzed that while existing R&D businesses are deeply embedded in Nvidia's product system, future investments such as new parks are more susceptible to management changes and market downturns.
Huang Renxun's departure has given the Israeli branch sufficient reason to worry that the original expansion plan may be weakened, as the Democratic Party, which will take over both houses of the US Congress in January 2027, has been quite dissatisfied with Israel's actions throughout the Middle East. Nvidia, no longer led by Huang Renxun, may no longer give Israel great development priority.
2、 Chip major buyer
Unlike Israel, although Gulf countries have also invested heavily in their cooperation with Nvidia, more returns depend on financing, construction, and computing power sales. Both sides need stable cooperation, and the chips available in the face of changes are not the same.
The equipment obtained by Saudi Arabia and the United Arab Emirates is at the forefront of Nvidia's product line.
In May 2025, HUMAIN, a subsidiary of Saudi Public Investment Fund, announced a partnership with Nvidia to deploy 18000 GB300 Grace Blackwell products in the first phase. The United Arab Emirates' Stargate is planning to build a 1-gigawatt computing facility located within a larger 5-gigawatt campus, with participation from G42, OpenAI, Oracle, Nvidia, and others.
In November of the same year, the US Department of Commerce approved G42 and HUMAIN to each purchase up to 35000 advanced chips equivalent to GB300, with accompanying security and reporting requirements. Project planning, approved quota, and actual production are several stages that need to be checked separately.
Huang Renxun is an important promoter of "sovereign AI". He emphasized that countries should use their own data, language, and industry needs to build their own AI capabilities. For Gulf countries hoping to convert energy revenue into new industries, this proposition is quite attractive.
But owning land, electricity, and servers can only solve part of the problem. The equipment requires continuous software and maintenance support, and the sale of computing power to external parties must also meet relevant export and end-user requirements. Huang Renxun can coordinate business cooperation, but cannot guarantee that these conditions will never change for the US government.
The San Francisco Declaration on November 15th further proposed the joint protection of sensitive technologies and critical infrastructure. The United Arab Emirates, Qatar, and Israel are on the list of ten signatory countries, and Saudi Arabia is not on this list. The declaration does not automatically impose an export ban at least at this stage, but Saudi Arabia's failure to sign the list may represent their wait-and-see attitude.
The remote computing power rules announced on the 16th directly touch on which customers data centers can serve. According to the publicly displayed text, providing advanced computing power beyond the threshold to entities with headquarters or ultimate parent companies located in specific restricted areas, and related equipment transactions and usage arrangements will involve licensing; If the stock arrangement that meets the conditions needs to continue, permission must also be obtained within 90 days after it takes effect. Leaving chips in the Gulf does not automatically exempt their users from US regulation.
For Gulf sovereign funds and local enterprises participating in the investment, this will change the basic conditions for investment establishment. The project initially calculated returns based on potential customer size, computing power prices, and equipment utilization. If the customer scope is limited, the original revenue forecast needs to be redone. Land, electricity, and computer rooms have already been invested, but customers may not be able to access them as planned. Gulf funds bear the risk of already incurring costs but being unable to fully operate the assets.
If the computer room is vacant, the loss will not stop at earning a small service fee. The agreed venue fees, electricity capacity fees, equipment maintenance, and debt interest still need to be paid. The utilization rate continues to be lower than expected, first compressing shareholder returns, and then potentially affecting debt repayment ability. In severe cases, the investing party may need to increase funds, accept asset impairment, or even be unable to recover the initial investment. National strategic projects also require cash flow, and the background of sovereign funds cannot eliminate losses.
Waiting for additional approvals from the US government also incurs costs. The depreciation of GPU technology will not be suspended because the license has not been obtained yet. During the project delay, new products with higher performance and lower energy consumption may have already entered the market. Even if later approved, the old equipment may have to lower its price to attract customers, and the originally planned investment payback period may be extended.
Therefore, Gulf countries need to reassess the actual autonomy of 'sovereign AI'. They have computer rooms and equipment, but may not be able to independently determine the entire customer scope; Having taken on long-term construction responsibilities, the operating conditions may change with external policies. The gap between asset ownership and operational decision-making directly affects how much these investments are worth.
In the upcoming cooperation negotiations, Gulf investors will definitely demand that the US government clearly state its responsibilities after policy changes. Whether to postpone procurement, adjust usage, or exit when unable to obtain permission, who will bear the losses when computing power customers cannot access, and whether there is a minimum procurement commitment or other guarantees should be included in the negotiation agenda of both parties. If not, the Gulf investors may bear the triple risks of construction, technological iteration, and customer access simultaneously.
3、 Should Gulf funds be withdrawn in line with the trend
The growth in demand for chips does not guarantee that stocks will continue to rise. Overvaluation takes into account future orders and profits in advance, and as long as the growth rate is lower than expected, investors will lower their quotes. For Gulf investors who have already invested heavily, what needs to be reconsidered at this point is how much return can be obtained by continuing to invest.
The direction for Saudi Arabia to reduce its overseas deployment has been clarified. The head of the Public Investment Fund proposed this year that local investment should account for about 80% and international investment should account for about 20%, lower than the previous high of about 30%. Domestic industrial construction has been given higher priority, and the funding share that overseas projects need to compete for has decreased accordingly.
This will directly change the conditions for US projects to seek funding from Saudi Arabia. In the past, as long as we were optimistic about the long-term growth of the US technology industry, there was reason to expand our allocation; In the future, the same amount of money will compete with Saudi Arabia's domestic manufacturing, energy, logistics, and urban construction industries. If American AI assets face valuation cuts, license tightening, and delayed payments at the same time, attracting new capital requires more convincing returns.
Saudi Arabia still retains significant US assets. The third quarter 13F filing submitted by the Saudi Public Investment Fund on November 16th showed a declared market value of approximately $29 billion. Existing investment continues to exist, and new capital is tilted towards the domestic market, both of which can occur simultaneously. The more important change for the US market is how much new Saudi funds can be obtained in the future.
The previous investment of Gulf funds has added weight to this change. According to Global SWF's mid year statistics, in the first half of 2026, Gulf sovereign funds committed approximately $53.9 billion in investments, involving 108 transactions, with nearly half flowing to the United States, with technology being an important direction. The US AI industry has become an important destination for these long-term funds.
The participation of the United Arab Emirates is particularly deep. In July, a consortium consisting of MGX, AI Infrastructure Partnership (AIP), and BlackRock's Global Infrastructure Partnership (GIP) completed the acquisition of Aligned Data Centers, with a transaction corresponding to approximately $40 billion in enterprise value and an additional commitment of $5 billion in growth capital. These investments directly link Gulf funds with the construction, leasing, and long-term returns of data centers in the United States.
The greater the investment, the more it is necessary to review whether the subsequent expansion is cost-effective. Signed investments need to be fulfilled, and private equity is difficult to exit quickly. Adjustments often occur first in the next round of follow-up investments, new projects, and additional budgets. The United Arab Emirates and other countries also have reasons to raise the investment threshold, requiring clearer customer contracts, regulatory licenses, and income arrangements before deciding whether to continue to increase.
Therefore, the way Gulf funds follow Saudi Arabia's adjustment may manifest as slowing down new investment, installment payments, and reducing expansion scale. The Huang Renxun incident has added reasons for re evaluating projects, and changes in export and computing power usage rules will enter into specific revenue calculations.
This adjustment is enough to affect the expansion of the US AI industry. Model companies and computing power projects that rely on external financing may need to postpone construction, reduce procurement, or accept more expensive financing if they cannot receive funding on time. The changes are subsequently transmitted to chip orders, hardware manufacturers' revenue expectations, and the valuation of listed companies.
What the US market needs to pay attention to is not just whether Gulf investors sell their existing stocks. Saudi Arabia has raised the priority of domestic investment, and other Gulf funds may also be more cautious. For the AI industry that relies on continuous expansion to support valuation, long-term investors who add less money may reschedule their original growth plans.
4、 Cost and Speed Reduction
The impact of Gulf capital on the US AI industry is mainly transmitted through corporate financing and infrastructure construction. Sovereign funds invest in model companies, and the funds are subsequently used to purchase cloud services; Invest in data centers, with funds used for land, electricity, servers, and network equipment. These expenditures ultimately generate revenue for American technology companies and support investors' expectations for the next phase of growth.
The AI infrastructure partnership between Abu Dhabi MGX and Kuwait Investment Authority connects Gulf funds with the US capital market. This platform aims to support up to $100 billion in investment with a target of $30 billion in equity capital, plus debt. Equity funds bear the initial risks of the project and provide a foundation for subsequent borrowing. For data centers built on this financing structure, if shareholders commit less money, it may affect the overall financing scale of the project.
Gulf investors now need to simultaneously review the risks on both ends. In the local market, for computing power projects that involve new licensing requirements, it is necessary to recheck the customer scope and revenue forecast; In the United States, model companies and data centers need to demonstrate that future revenue can cover increasingly high construction and financing costs. The same batch of funds must balance domestic projects and overseas returns, and US AI assets must provide sufficient returns in order to continue receiving additional investment.
The financing price has given a clear change. Reuters reported on the 17th that CoreWeave, a US cloud computing company that mainly purchases Nvidia GPUs to build computing power clusters, has seen the market yield of a bond increase from 9.625% at the time of issuance to nearly 15%, reflecting investors' demand for higher risk compensation. Nvidia also signed a procurement arrangement with the company with an initial value of 6.3 billion US dollars to purchase the remaining unsold computing power according to the agreed terms. This makes Nvidia both a chip supplier and an important supporter of customer business.
For Gulf investors, such arrangements require a clear understanding of where the final returns come from. No matter how many computer rooms are built, customers still need to continue paying in order to cover equipment, electricity, and borrowing costs. Suppliers provide support that can help expand projects, but cannot guarantee sufficient business for every data center.
If Gulf funds raise the investment threshold and reduce additional investment in American AI companies and infrastructure projects as a result, companies that rely on these funds will need to slow down construction and delay chip procurement. The impact subsequently spread to suppliers such as Nvidia, weakening expectations for new orders and revenue growth. Even if Gulf investors do not sell their existing US stocks and invest less in the next round, they may still put pressure on the valuation of US AI stocks through this chain.
Less follow-up investment in the next round, delayed approval of the new park by a few months, or requiring the implementation of customer contracts before funding can all change the project schedule. Large companies with sufficient cash flow are better able to withstand such adjustments, while computing power companies that rely on continuous financing need to reduce their scale or accept more expensive funds.
The cost of expanding security reviews in the United States will not be limited to Gulf data centers. Customer access restrictions affect project revenue, which in turn affects financing capacity, which in turn affects orders from US equipment suppliers. The United States hopes that Gulf capital will support its AI industry in the long run, so investors need to be able to judge who the approved projects can serve, how much money they can earn, and who will bear the losses after policy changes. The security rules are formulated by the United States, and the repricing of related investments will ultimately return to the US capital market.
AI connects Silicon Valley's chips, Israel's research and development, and Gulf's energy and capital into the same global industrial network.
Huang Renxun's resignation may soon fade out of the headlines, but the financing, procurement, and cooperation adjustments surrounding it will still spread outwards like ripples, entering investment plans and industrial layouts in different countries. The extent of the impact of this storm will depend on the upcoming changes in orders, funding, and policies. The world is becoming more interconnected due to AI, and it is becoming increasingly difficult for a key company's changes to stop at its own board of directors.
(The author is the director of the Middle East Research Institute at Pangu Think Tank)