-+ 0.00%
-+ 0.00%
-+ 0.00%

Data center “tax exemption bonuses” have changed! More than 10 states in the US are reviewing tax benefits and AI computing power expansion has been blocked

Zhitongcaijing·09/09/2026 13:49:10
Listen to the news

The Zhitong Finance App learned that tax benefits affecting the construction process of major data centers in the US are facing a re-review. More than 10 states are re-evaluating or reducing subsidies, while more than 35 states are still providing related benefits. Ohio's data center tax exemption last year exceeded 1.5 billion US dollars, more than 10 times the initial forecast, prompting some members of Congress to call for the cancellation of concessions and renegotiation of long-term agreements.

E-commerce and cloud computing giant Amazon, which is leading the unprecedented wave of AI data center construction in the US, emphasized that it has invested nearly 40 billion US dollars and created thousands of jobs in the state since 2015. The core of the dispute between US tech giants and state governments is how the government can further balance investment, employment, and tax benefits, as well as utility costs brought about by data center expansion; for tech giants, uncertainty about long-term tax arrangements is becoming a new variable in return on capital expenditure.

For Amazon, Meta, and Google, the direct impact of shrinking tax benefits is rising construction costs and pressure on post-tax returns for projects. After data center server clusters and related equipment lost sales tax exemptions, cash expenses for purchasing and upgrading equipment increased; if enterprises still need to bear more electricity supply and power grid construction costs, the payback period for data centers may also be extended. However, Ohio suspended new applications and proposed abolition of concessions and renegotiation of existing contracts with lawmakers. This is a trade-off measure at different stages. For the time being, it cannot be determined that tech giants have lost all long-term tax exemptions based on this.

The Astra release boosted the market's expectations for general artificial intelligence (AGI), leading to a stronger computing power demand trajectory. In particular, the new “pay-for-performance” growth model is expected to bring stronger demand for total computing power, while recent more direct evidence of computing power demand comes from the AI R&D process itself — that is, the “recursive self-improvement (RSI)” development trajectory where AI began to “create AI.”

As the advanced and cutting-edge model led by Astra brings more and more strong demand for AI computing power, Morgan Stanley expects the data center comprehensive capital expenditure of the four largest supercloud computing and AI application vendors in North America to rise from US$917 billion in 2026 to US$1.47 trillion in 2027 and US$1.64 trillion in 2028. The deployment capacity is expected to expand from 35 gigawatts in 2025 to 145 gigawatts in 2028.

Amazon, Meta, and Google face new risks as states re-examine tax credit agreements

More than a dozen states are reevaluating or reducing tax subsidies for data centers. Once popular as an important growth engine driving the accelerated expansion of the AI economy, these facilities are now increasingly viewed by local leaders as a political risk. As a result, as opposition against these facilities continues to heat up, America's largest tech giants, such as Amazon (AMZN.US), Facebook's parent company Meta (META.US), and Google (GOOGL.US), are at risk of losing decades of tax exemptions.

More than a decade ago, Ohio lawmakers exempted tech companies from sales tax on the purchase of computer servers and other equipment needed for data centers, betting that these multi-million dollar tax adjustments would attract much-needed investment to the state. The move worked, making Ohio one of the main locations for data centers. However, the report said that the AI boom that followed dramatically expanded the scale of tax exemptions, exceeding 1.5 billion US dollars last year, more than 10 times the state's initial estimate.

Today, some state lawmakers, including Democratic Representative Tristan Reed, are desperate to abolish sales tax exemptions and renegotiate previous agreements with technology companies such as Amazon, Meta Platforms, and Alphabet's Google. These businesses have been granted tax exemptions for decades by signing contracts with state governments. The report said that in many states, tech giants already enjoy more than 1 billion US dollars in tax exemptions every year.

“They seem richer than God, and even without these incentives, they have these facilities to build capacity,” Red said. He is proposing a new tax on data centers and requiring developers to pay more for electricity and power infrastructure. His constituency covers parts of Cleveland, and local residents have expressed opposition to these facilities. He hoped these new pressure measures would force companies to sit down at the negotiation table.

Amazon said in a statement that since 2015, the company has invested nearly $40 billion in data centers in Ohio, created thousands of jobs, and paid nearly $11 million in real estate taxes and related expenses in the state last year. Meta and Google declined to comment, as did Governor Devin's office.

The offensive against preferential tax provisions is the latest example of a wave of opposition to AI across the US. According to the report, Missouri City Councilor John Perkins previously voted to approve the provision of multi-billion dollar tax benefits to a data center, but then lost in the recent election.

Despite recent polls showing that the vast majority of Americans don't want data centers built near them, President Trump is urging voters to welcome these facilities. “If you want your state or town to be rich, pay less taxes, create huge wealth, and want your house and other assets to be more valuable, then you want a large AI data center,” he said on Friday. “If you want to live in poverty, crime, and mess, then I would say don't approve the construction of data centers.”

According to media reports, Ian Bocaccio, a senior researcher from the tax services company Ryan, said that the investment attractiveness of Ohio, Arizona, and Illinois declined after showing their intention to suspend or completely cancel some tax exemptions. He said he is urging data center customers to consider taking advantage of other investment tax benefits offered to businesses by various states. Industry observers generally say that at the end of the day, the actual benefits and future growth benefits of these facilities are too many for the government to shut them out.

How to influence the global AI computing power investment theme?

Data center boycotts are expanding restrictions on AI investment from chip and power supply to construction permits and community acceptance. Annenberg announced in August. The actual survey conducted from June to July showed that the proportion of adults opposed to building local data centers rose from 49% to 61%, but there was no significant change in public opinion on the overall impact of AI. Meanwhile, Data Center Watch confirmed that at least 75 projects worth about $130 billion were blocked or postponed in the first quarter. As far as the AI industry chain is concerned, the direct impact is that the time for new computing power to be put into operation has become more uncertain, and equipment delivery, cloud service expansion, and confirmation of AI-related revenue generation that the market focuses on may be delayed as a result of the AI computing power industry chain falling back due to reduced valuations.

The economic essence of the dispute is who actually bears the costs of additional electricity generation, transmission, and water supply. After technology companies signed the “Electricity Bill Payer Protection Pledge,” there are still differences in cost sharing. Some media reported in early September that Microsoft is appealing against Virginia's new regulations requiring data center developers to bear the cost of transmission infrastructure construction in advance.

The case described above shows that “enterprises bear their own electricity costs” also involves the timing of payment, the scope of infrastructure, and risk allocation during implementation. If more expenses need to be borne in advance by the developer, the initial capital investment and financing requirements for the project will increase; if the costs are passed on to residents, it may continue to cause political resistance. Promises how to include electricity price agreements and construction conditions will directly affect the return on investment in data centers.

As far as the AI computing power investment theme is concerned, the first impact of policy changes is where to build, how high the cost of construction is, and when the return on investment is formed. For example, data center discounts in Ohio in the US cover computing equipment, cooling systems, power equipment, and some construction materials, so the scope of adjustment may span the entire data center campus. According to this cost transmission mechanism, Amazon AWS and Google Cloud need to re-measure the expansion costs of marketable computing power, while Meta faces changes in infrastructure costs to support advertising recommendations and AI products. However, capital expenditure may not have been reduced but may be redistributed between regions, and it is important to note that rising tax burdens alone do not prove a decline in demand for terminal AI.

The potential positive impact in the medium to long term depends on whether additional costs can be exchanged for more clear construction and access conditions — if the project can obtain stable power supply arrangements after cost sharing is clear, it will be easier for companies to plan and start production and form a timeline for strong AI-related revenue generation. AEP Ohio has implemented a data center specific electricity price and back-meter power grid system, and is advancing new projects through batch research, signing contracts, and clarifying estimated power supply dates. This shows to a certain extent that technology companies assume more infrastructure responsibilities and the continued expansion of projects can occur at the same time.