Amazon Aims for AI Supremacy With $8B Data Surge in Ohio

Just as it did in 2021—when it doubled the size of its logistics network to more than 450K SF to address the surge in e-commerce during the pandemic—Amazon is marshalling its unmatched financial resources to double the capacity of its data center network in order to put its cloud platform at the center of generative AI.

Amazon Web Services (AWS) aims to be the leading platform for training large language models (LLMs) for generative AI applications, directly challenging Microsoft—for now, the industry leader with OpenAI’s GPT 4—as well as Google and Salesforce—and, coming soon, initiatives from mixed-martial-arts enthusiasts Elon Musk and Mark Zuckerberg—in a tech-sector version of an arms race between superpowers.

Data centers around the world are racing to expand their capacity with servers equipped with AI-programmable nano-scale semiconductors like the chips pioneered by Nvidia that enabled the LLM breakthroughs of the past year.

AWS announced on its website on Monday that the tech giant will be spending $7.8B to expand its data center capacity in Ohio between now and the end of the decade.

The Ohio expansion—which will nearly double the $6.3B worth of investments Amazon already has made building hyperscale facilities in Franklin and Licking counties—is the second-largest economic development commitment in the history of the Buckeye State, behind Intel’s announcement last year that it will be investing $20B in new semiconductor fabs in Ohio.

Amazon’s total planned investment in Ohio data centers is only half the size of an expansion AWS announced earlier this year it will be undertaking in Virginia.

In January, Amazon followed through on a plan to shift capital expenditures from its overextended logistics network to the AWS data infrastructure by committing to invest $35B by 2040 in multiple new data campuses that will exponentially expand its cloud capacity in Virginia, home of the world’s largest data center cluster.

The Virginia Economic Development Partnership (VEDP) sealed the deal with AWS with a new Mega Data Center Incentive Program. Under the program, Amazon will receive a 15-year extension of data center sales and use tax exemptions on qualified equipment and enabling software, and an MEI performance grant of $140M for site and infrastructure improvements, workforce development and other project-related costs.

Last month, Amazon announced plans to demolish up to a dozen suburban office buildings in Loudoun and Fairfax counties and replace them will new server farms. The cloud-computing leader will replace the office campuses with five data centers encompassing nearly 1M SF, the Washington Business Journal reported.

As the tech giants ramp up their generative AI arms race, governments around the world—or, at least, in the EU and the US—are racing to install some regulatory guardrails for the rapid development of LLMs.

On June 14, the European Parliament voted to impose a comprehensive regulatory regime on generative AI. The EU’s AI Act sets transparency requirements for LLMs like GPT-4 that including disclosing all content generated by AI, publishing summaries of copyrighted data used for LLM training and designing each model to prevent it from generating illegal content.

Under the AI Act, AI systems trained in eight specific sectors will require registration in an EU database.

These include biometric identification and categorization of natural persons; management and operation of critical infrastructure; education and vocational training; employment, worker management and access to self-employment; access to and enjoyment of essential private services and public services and benefits; law enforcement; migration, asylum and border control management; and assistance in legal interpretation and application of the law. The EU’s AI Act also bans what it calls “unacceptable risk” AI systems that it considers a threat to people.

These include cognitive behavioral manipulation of people or specific vulnerable group, citing as an example voice-activated toys that encourage dangerous behavior in children; “social scoring,” which it defines as classifying people based on behavior, socio-economic issues or personal characteristics; and real-time and remote biometric identification systems, including facial recognition.

In the US, President Biden has been holding a series of meetings with the leading developers of generative AI to discuss a regulatory framework. Sen. Chuck Schumer, the Senate majority leader, recently announced he will soon introduce a comprehensive regulatory package covering generative AI.

Meanwhile, Microsoft appears to have found a creative way to selectively introduce GPT-5, which will be able to perfectly mimic human voices and match them to deep-fake videos: Paul McCartney announced he’s been able complete a new Beatles song by using generative AI to recreate the late John Lennon’s voice.

No word on whether Sir Paul, who once famously tried to reverse the songwriting credit on several Beatles tunes from “Lennon-McCartney” to “McCartney-Lennon,” will ask the bot to replicate Eric Clapton filling in for George Harrison on lead guitar.