The DeepSeek announcement continues to reverberate throughout the world. A timely summit held in Paris this week underscores DeepSeek's impact, with French President Emmanuel Macron working to elevate his country and the EU to a level competitive with the US and China in the new AI-development race.

There were more than 100 nations represented at this AI Action Summit, with overarching themes of the future of work, the public interest, innovation and culture, trust and global AI governance. One statement from the French government stated that “The time has come to develop open, universally accessible technical solutions and work on common standards to avoid fragmentation and encourage convergence around high levels of safety and security for artificial intelligence.”

Coincidentally, taking a business-first perspective, a CIO-focused event in Silicon Valley produced by The Wall Street Journal noted that ROI “from generative AI continues to be elusive,” while asking whether corporate technology leaders “need to play a more significant role in AI governance and ethics, data governance, and risk management.”

The latter event also unleashed a quantum computing genie, with startup co-founder Pete Shadbolt of PsiQuantum claiming his company will be delivering quantum computing in two years, in contrast to the perpetual five-year wait that's been the quantum norm for several years. A separate survey found that most C-level execs are simply too focused on the present, AI-driven challenges to be overly alarmed about quantum's implications, and Shadbolt also worked to dampen fears of it by noting that quantum-driven security should be able to withstand quantum-based cyberattacks.

As far as actual AI center construction, the Big Three cloud vendors are sticking to commitments of more than $250 billion over the next few years, the Stargate $500 billion vision remains in place, a 3GW desert vision in Rajasthan, India is in the works (at an estimated cost of at least $300 billion), to list just three of the megadreams being discussed. It thus becomes easy to envisage more than a $1 trillion in data center investment being discussed at the moment.

The world's entire existing data center footprint would cost $400 to $500 billion to replicate, so $1 trillion and more may seem optimistic. But even a moderate projection by a major real-estate/data center developer last year projected more than a doubling of the existing footprint within four years.

So $1 trillion and more could be the low end of projections as 2025 moves forward; initial panic caused by the DeepSeek announcement that questioned the need for massive new compute resources has been replaced by the time-tested notion that if a product becomes dramatically less expensive (ie telecommunications, consumer electronics, or AI software) then demand will spike even more dramatically. This notion seems to be driving more enthusiasm than before for massive AI center development, rather than less.