OpenAI is simultaneously advancing two strategic pillars: AI infrastructure and enterprise software. It is building a data center in Georgia with a total investment exceeding USD 30 billion and a power capacity of 3.2 gigawatts, with the first phase of power expected to come online in 2028. Concurrently, OpenAI has launched Presence, an enterprise AI agent platform that deeply integrates AI models with enterprise data and business workflows, offering automated solutions for functions such as customer service and sales.
OpenAI is simultaneously intensifying its strategic investments in AI infrastructure and enterprise software. The developer of ChatGPT announced plans to build a hyperscale data center in Georgia, USA, with a total investment exceeding $30 billion and a planned power capacity of 3.2 gigawatts, and to launch Presence, an enterprise AI agent platform, further expanding into the enterprise software market.
On July 22, according to Bloomberg, OpenAI has signed a power supply agreement with Georgia Power to secure 3.2 gigawatts of electricity capacity for its data center in Georgia. The first several hundred megawatts are expected to come online in 2028, with the entire project scheduled for completion by 2032. Sachin Katti, Vice President of Compute Strategy, stated that if built at full scale, the total investment would exceed $30 billion, and the company is currently seeking partners for co-development.
Meanwhile, OpenAI has officially launched Presence, its enterprise AI agent platform, which deeply integrates AI models with internal enterprise data, policies, and business workflows to deliver automated solutions for use cases such as customer service, sales, and IT support—moving beyond merely offering model access.
As competition in foundational models intensifies and pricing continues to decline, OpenAI is simultaneously scaling up its compute capacity investments and extending its reach into enterprise software services, building new competitive advantages through coordinated strategies across both infrastructure and applications.
Building a 3.2 GW Data Center with Over $30 Billion Investment
The Georgia project will become one of OpenAI’s largest AI infrastructure investments to date.
Under the agreement, OpenAI will secure 3.2 gigawatts of power capacity, with the initial several hundred megawatts expected to be deployed incrementally starting in 2028. The entire construction timeline will extend through 2032. A 3.2-gigawatt power capacity is equivalent to the electricity demand of approximately 2.4 million U.S. households.
Katti noted that if the project is developed at full scale, the total investment will surpass $30 billion. OpenAI has already committed to investing $20 billion in Effingham County, Georgia, in exchange for a pending package of tax incentives and other benefits. Simultaneously, the company is in discussions with potential partners regarding financing and joint development, and it has not ruled out independently operating certain facilities.
According to The Wall Street Journal, amid surging demand for AI computing power, OpenAI has revised its projected future cloud infrastructure spending upward to $750 billion, signaling its continued commitment to significantly increasing long-term capital expenditures.
From Participant to Leader in 'Stargate'
The Georgia project also reflects OpenAI's strategic realignment of Stargate.
Stargate was initially launched jointly by OpenAI, Oracle, and SoftBank, but as the project progressed, collaboration models across campuses have gradually diversified, with OpenAI taking on an increasing share of planning and design responsibilities.
Katti stated that after two or three generations of AI infrastructure development, the company has accumulated substantial design experience, enabling further optimization of data center architecture to reduce costs and accelerate construction timelines—hence its decision to lead overall design in more projects.
OpenAI also plans to progressively open-source its AI data center design standards, including specifications for data centers, servers, and network architectures, aiming to build an open ecosystem similar to Meta’s Open Compute Project.
Meanwhile, the company is also adjusting its global investment footprint. According to reports, OpenAI has recently exited certain data center projects in the UK and Norway and paused expansion of its flagship Stargate campus in Texas, redirecting resources toward large-scale infrastructure initiatives of greater strategic importance.
Proactively Addressing Community Concerns
The rapid expansion of large-scale data centers has also sparked growing concerns regarding power consumption, water usage, and environmental impact.
In response, OpenAI has proactively introduced several accompanying measures. The company stated it will bear the full cost of building associated power transmission and distribution infrastructure, ensuring no related expenses are passed on to local residents; data centers will voluntarily curtail electricity use during periods of high grid demand and employ closed-loop recirculating water cooling systems to minimize water consumption.
Additionally, the company pledged to pay hundreds of millions of dollars in taxes to state and local governments in the future, establish a community fund, and provide local students with access credits to OpenAI products.
Chris Lehane, OpenAI’s Chief Global Affairs Officer, stated that the company will begin ongoing engagement with local residents this week and commission independent third parties to conduct oversight audits, ensuring the project fully addresses community concerns.
Presence targets the enterprise software market
While continuing to expand its infrastructure, OpenAI is also accelerating its enterprise product strategy.
The newly launched Presence does not merely offer model APIs; instead, it deeply integrates AI models with a company’s internal data, policies, existing software, and business workflows, enabling enterprises to deploy AI agents capable of autonomously completing tasks.
The platform primarily serves use cases such as customer service, sales, and IT support, and can automatically handle specific business processes including billing inquiries, insurance claims, and employee IT service requests.
Compared to simply selling access to models, such enterprise software products offer higher customer stickiness and help OpenAI establish a more stable, long-term revenue stream. Currently, Presence is deployed through a customized implementation model, requiring enterprises to collaborate with OpenAI engineers or designated partners for deployment. OpenAI has disclosed that it has already begun using the platform internally for its own English-language customer service hotline.
As competitors—including Anthropic, Google, Meta, xAI, Mistral, and DeepSeek—continue to enhance model capabilities and drive down pricing, foundational models are increasingly becoming commoditized. For OpenAI, deeply embedding AI into enterprise workflows and delivering end-to-end solutions—rather than merely selling model access—is emerging as a key strategic focus for its next phase of competition.
Editor/KOKO