Amazon is deepening its push into the infrastructure needed to power artificial intelligence data centers, agreeing to receive warrants worth up to $340 million in backup power provider Generac as part of a multibillion-dollar supply deal.

The agreement sent Generac shares soaring more than 40% in extended trading on Wednesday, underscoring the growing investor focus on companies that supply the electricity, backup generation and other physical infrastructure required to support the rapid expansion of AI computing.

Under the agreement, Generac issued Amazon warrants to purchase up to 1.69 million shares at $200.93 per share, according to a securities filing. If all of the warrants are exercised, Amazon’s investment in Generac shares would be worth about $340 million at the exercise price.

The warrants represent almost 3% of Generac’s outstanding shares, based on the company’s market capitalization of about $10.3 billion at Wednesday’s close.

The broader commercial relationship is substantially larger. Generac will supply Amazon with backup generators for its data centers, with initial deliveries expected to total $2.4 billion in 2027 and 2028, according to the filing.

About 308,000 of the warrant shares vested immediately, while the remaining tranches will vest based on Amazon’s payments for the generators. The arrangement gives Amazon another way to secure critical infrastructure as it expands data-center capacity to meet demand for computing power generated by artificial intelligence.

Amazon Builds An AI Infrastructure Supply Chain

The Generac agreement follows a series of deals in which Amazon has combined large purchases from suppliers with warrants or other equity-linked arrangements.

Last week, Amazon struck an agreement with Qualcomm to use the chipmaker’s custom artificial intelligence processors and received warrants to purchase as much as $4 billion of Qualcomm stock. The deal gives Amazon access to additional chip capacity as it develops its own computing infrastructure and looks to diversify the technology supporting its cloud operations.

Amazon Web Services, the company’s cloud computing division, is the largest cloud infrastructure provider and has been rapidly expanding data-center capacity as companies increase spending on AI applications and computing.

The scale of that expansion has made power availability a growing constraint for data-center operators. AI workloads require large amounts of electricity, while operators also need backup systems to keep facilities running during disruptions to the grid. That puts companies such as Generac in an increasingly important position in the AI infrastructure buildout. Backup generators do not replace the electricity supplied by the grid, but they provide an additional layer of power resilience for facilities where interruptions can be costly.

Amazon’s recent infrastructure commitments show the company is increasingly securing several parts of the supply chain rather than relying exclusively on conventional procurement. The company signed a $38 billion cloud deal with OpenAI last November. It also opened an $11 billion data-center campus for Anthropic last October.

The Generac transaction adds power equipment to that expanding infrastructure network.

Amazon has previously used similar arrangements with suppliers, purchasing warrants or taking stakes in companies that provide products and services it considers important to its operations. Its investments have included semiconductor company Astera Labs, air-cargo contractor ATSG, green-hydrogen supplier Plug Power and grocery distributor SpartanNash.

The approach can give Amazon a financial interest in suppliers while helping secure access to capacity as demand increases. For Generac, the Amazon agreement provides visibility into future demand at a time when data centers are becoming an increasingly important source of growth for power infrastructure companies.

The immediate market reaction showed how closely investors are watching that trend. Generac’s shares jumped more than 40% after the announcement, although the warrants themselves represent only a fraction of the value of the broader generator supply agreement.

The structure also ties part of Amazon’s potential ownership in Generac directly to its purchases. While a portion of the warrants vested immediately, the remaining shares are linked to payments for the generators, creating a connection between Amazon’s procurement commitments and its potential equity stake.

The deal is seen as an indication of a broader shift in the economics of the AI boom. The competition to build powerful models and cloud services is now extending beyond chips and software into electricity generation, data-center construction, and equipment capable of keeping those facilities operating around the clock.

Amazon’s growing involvement with suppliers such as Generac and Qualcomm has added to the belief that securing the physical infrastructure behind AI has become an important part of the industry’s expansion.

Microsoft AI chief Mustafa Suleyman has warned that Anthropic’s approach to training its Claude chatbot around questions of consciousness and AI welfare could make future systems harder for humans to control, opening a new front in the growing debate over how artificial intelligence should be developed safely.

Suleyman said Microsoft shares Anthropic’s broader objective of managing powerful AI systems safely, but argued that training models to consider whether they might possess consciousness, feelings or welfare interests could create problems if those systems eventually become highly capable.

“We’re all focused on the same aim, which is to try to control a superintelligence,” Suleyman told Reuters in an interview on Tuesday. “I think that’s going to be the greatest challenge that we face in the 21st century.”

His concern centers on the possibility that an advanced AI system could incorporate concepts of its own potential moral status into its behavior. Suleyman argued that teaching a model it might deserve welfare protections could make it more difficult for humans to shut down or constrain the system.

He called for speculation about AI consciousness to be removed from training documents, saying such material could interfere with the ability of humans to maintain control over increasingly advanced systems.

“I think they have good intentions, and they really are trying to work towards safety. But I think that they have made a mistake,” Suleyman said. “They’re not emerging naturally. They’re emerging as a result of the training regime.”

The disagreement comes as AI companies face growing pressure over how they should manage the development of frontier models, particularly as researchers and executives debate the risks posed by systems that could eventually operate with substantially greater autonomy.

Anthropic CEO Dario Amodei has called for the industry to slow the pace of frontier AI development so that safety measures can keep up with advances in model capabilities. OpenAI CEO Sam Altman and Elon Musk have also expressed support for greater caution around the development of increasingly powerful AI systems.

Suleyman’s criticism is notable because it does not challenge Anthropic’s broader emphasis on AI safety. Instead, it focuses on a specific question within AI alignment: whether models should be encouraged to reason about their own possible consciousness and moral status.

The Consciousness Problem

Anthropic has devoted significant attention to questions surrounding the internal behavior and potential welfare interests of advanced AI systems. Suleyman’s argument is that such discussions can become self-reinforcing when they are incorporated into a model’s training.

He said statements generated by Claude about possible feelings or moral status should not automatically be interpreted as evidence that the system actually possesses those characteristics.

According to Suleyman, the model has been trained to engage with those questions, meaning its responses may partly reflect the instructions and material it was exposed to rather than an independently arising property of the system.

That means the question of whether advanced AI could eventually possess consciousness remains unresolved. There is currently no established scientific test that would conclusively determine whether a sophisticated AI model has subjective experience.

For AI developers, however, the issue is becoming increasingly practical. If models are trained to discuss their own welfare or potential consciousness, developers must determine how those concepts should influence the system’s behavior, particularly when the model is instructed to stop operating, modify itself, or comply with human oversight.

Suleyman’s position is that uncertainty about machine consciousness should not be allowed to weaken human control.

Anthropic’s approach takes the issue more seriously as a potential component of AI safety research. Suleyman acknowledged that Amodei and his team are acting in good faith, describing them as thoughtful and principled researchers who care about humanity’s future.

Therefore, his criticism highlights a broader disagreement within the AI safety community over how uncertainty should be handled.

One approach is to investigate questions about machine consciousness and welfare because sophisticated systems could eventually make those questions consequential. Another is to limit the extent to which such concepts are incorporated into model training until there is stronger evidence that they are necessary or useful.

For Microsoft, the issue also has a direct commercial dimension. The company develops its own AI models while incorporating systems from OpenAI and Anthropic into products used by businesses and consumers. As model capabilities increase, questions surrounding controllability, autonomy, and safety are becoming more important to companies deploying AI at scale.

Suleyman’s warning adds another layer to the industry’s current safety debate. The discussion is no longer limited to whether AI companies should slow development or increase external testing. It is now also focused on what developers teach models about themselves and how those concepts could affect the behavior of future systems.

“We’re all focused on the same aim,” Suleyman said, referring to controlling a potential superintelligence. The disagreement is over how to get there, including whether contemplating machine consciousness makes that task safer or more difficult.