Global investments in artificial intelligence are yielding remarkable figures, with approximately $3 trillion (£2.3 trillion) allocated to data centers.
These immense facilities serve as the backbone for AI applications like OpenAI’s ChatGPT and Google’s Veo 3, driving the training and functioning of technologies that have attracted billions from investors.
Although there are worries that the AI boom might lead to a bubble poised to burst, indicators of such a downturn are currently absent. Recently, Nvidia, a Silicon Valley AI chip manufacturer, became the first company to reach a valuation of $5 trillion, while Microsoft and Apple each hit a $4 trillion valuation for the first time, marking a historic moment. OpenAI’s restructuring now appraises it at $500 billion, with Microsoft’s investment exceeding $100 billion. Projections suggest a potential $1 trillion surge as early as next year.
Moreover, Google’s parent company Alphabet announced $100 billion in revenue for a single quarter, driven by an increasing demand for AI infrastructure. Apple and Amazon also recently reported robust results.
Trust in AI extends beyond the financial sector; local communities housing the AI infrastructure are equally invested.
In the 19th century, the demand for coal and steel determined Newport’s trajectory. Today, Welsh towns are looking forward to a fresh era of growth generated by the latest global economic transformation.
At the site of a former radiator factory on the outskirts of Newport, Microsoft is constructing a data center to cater to the tech industry’s increasing demand for AI.
While standing on the concrete floor where thousands of buzzing servers will soon be installed, Dimitri Batrouni, the Labour leader of Newport City Council, remarked that the Imperial Park data center represents an opportunity to delve into the economy of the future.
“In a city like mine, what should we do? Should we dwell on the past in hopes of reviving the steel industry and bringing back 10,000 jobs? That’s not feasible. Or should we embrace the future?” he stated.
Yet, despite the current optimistic outlook regarding AI, uncertainties linger concerning the sustainability of spending in the tech sector.
The top four players in the AI industry (Amazon, Meta, Google, and Microsoft) are ramping up their AI spending. Over the upcoming two years, they are expected to invest more than $750 billion in AI-related capital expenditures, covering not just data centers and staff, but also the chips and servers they contain.
This expenditure is highlighted by the American investment firm Manning & Napier, which describes it as “nothing too remarkable.” The Newport facility alone could demand hundreds of millions of dollars. Recently, Equinix, based in California, announced intentions to invest £4 billion in a central hub in Hertfordshire.
Joe Tsai, chairman of the Chinese e-commerce giant Alibaba, cautioned in March that the data center market was beginning to exhibit signs of oversupply. “We’re starting to observe the early stages of a potential bubble,” he commented, referencing projects that finance constructions without securing commitments from prospective clients.
There are already 11,000 data centers globally, representing a 500% increase over the past two decades, and more are on the horizon. The means of funding this expansion raises concerns.
Analysts from Morgan Stanley predict that worldwide spending on data centers will approach $3 trillion by 2028, with $1.4 trillion of that anticipated from cash flow generated by large US tech firms known as “hyperscalers.”
Consequently, $1.5 trillion will need to be sourced from alternative means, such as private credit, which has been increasingly scrutinized by institutions like the Bank of England. Morgan Stanley estimates that private credit could cover more than half of the funding shortfall. Meta Inc. utilized private credit markets to raise $29 billion for an expansion of a data center in Louisiana.
Gil Luria, the head of technology research at DA Davidson, described investments in hyperscalers as a “healthy” aspect of the current boom, while labeling the remainder as “speculative assets devoid of customers.”
He noted that the debt being utilized could lead to repercussions extending beyond the tech sector if the situation deteriorates.
“Providers of this debt are so eager to invest in AI that they may not have adequately assessed the risks associated with a new and unproven category reliant on assets that depreciate quickly,” he indicated.
“We are in the initial phase of this influx of debt capital, but if it escalates to hundreds of billions of dollars, it could ultimately present structural risks to the global economy.”
Hedge fund founder Harris Kupperman noted in an August blog that data centers: depreciate at twice the rate of revenue generation.
Supporting this expenditure are heightened revenue forecasts from Morgan Stanley, which estimates that income generated from AI innovations such as chatbots, AI agents, and image generators could grow to $1 trillion by 2028 from $45 billion last year. To substantiate these revenue projections, tech firms are counting on enterprises, the public sector, and individual users to generate sufficient demand for AI and fund it.
OpenAI’s ChatGPT, a landmark product of the AI wave, currently boasts 800 million weekly active users. This statistic is a boon for optimists. However, concerns have arisen regarding user acquisition. For instance, investor confidence in the AI surge took a hit in August when the Massachusetts Institute of Technology released a study indicating that 95% of organizations reported zero return on investment from generative AI projects.
According to the Uptime Institute, which inspects and evaluates data centers, many projects go unconstructed, suggesting that some are part of a hype cycle and fail to materialize.
“It is crucial to understand that much of this is speculative,” stated Andy Lawrence, the Uptime Institute’s executive director of research. “Frequently, many data centers announced with great excitement are either never built or are only partially constructed and developed progressively over a ten-year span.”
He further added that numerous data centers unveiled as part of this multitrillion-dollar initiative “will be specifically designed for or primarily intended to support AI workloads.”
Microsoft has pointed out that its Newport data center will not solely serve AI. Data centers form the core for AI systems like ChatGPT and Microsoft’s Copilot but also cater to everyday IT tasks many take for granted (like managing email traffic, storing company files, and supporting Zoom calls) as providers of “cloud” services, where companies lease servers rather than purchasing them outright.
“The infrastructure has multiple applications, making it highly versatile,” explained Alistair Speirs, general manager of Microsoft’s cloud operations.
However, various large-scale projects are completely committing to AI. The US Stargate initiative is a $500 billion partnership among OpenAI, Oracle, and SoftBank, with plans to establish a network of AI data centers throughout the U.S. A British counterpart will also be set up in North Tyneside, in the northeast of England. Microsoft is constructing the most powerful AI data center in Fairview, Wisconsin, and is backing a dedicated AI site in Laughton, Essex, while Elon Musk’s xAI is developing a colossal project in Memphis, Tennessee.
Construction of an estimated 10GW of new data center capacity worldwide—equivalent to around a third of the UK’s electricity demand—is expected to commence this year, as reported by the property group JLL. However, this represents total maximum capacity, as data centers generally operate around 60% of their capacity.
JLL projects another 7GW will be completed this year.
The growth rate is swift, with current global data center capacity standing at 59GW, and Goldman Sachs forecasting capacity will double by the end of 2030. This expansion will elevate the costs related to the infrastructure, necessitating $720 billion in grid investments to satisfy that energy demand, according to Goldman.
Mike O’Connell, a construction safety specialist from Newport, has returned as a consultant at the Newport facility. With a career spanning oil rigs, offshore wind farms, and data centers globally, he returned to his hometown, now a tech hub filled with data centers and semiconductor firms.
“My aim is to remain within my local community,” he stated. Mr. O’Connell’s teenage grandson is embarking on his career at the Newport site as an electrical apprentice. There is optimism that such a data center will offer generational employment opportunities for the area.
Investors and tech giants are committing trillions of dollars in investments with hopes for long-term returns.
Source: www.theguardian.com
