SoftBank’s €75 billion pledge masks French pain

France staged its ninth Choose France summit with the usual polish: investors at Versailles, ministers selling stability, and SoftBank promising "up to" €75 billion for AI data centres. That pushed cumulative announcements past €100 billion since 2018.

Two days earlier, the national statistics office had published the less flattering picture. French GDP contracted -0.1% in Q1 2026, the first fall since mid-2020. The country presenting itself as Europe’s investment magnet is, for now, going backwards.

There is no public scorecard for the pledges. After nine summits and more than 230 announced projects, the government has never published a systematic account of how many were actually built. Some clearly were. The problem is that nobody outside the system knows how many.

The figures that refuse to line up

France lost a net 20,000 industrial jobs in 2025. Manufacturing’s share of GDP slipped from 10% in 2017 to 9.5% in 2025. Unemployment climbed back to 8.1%.

Olivier Lluansi, a former industry adviser to the Élysée Palace, put a hard number on the gap. Industrial job creation averaged 15,000 a year under Emmanuel Macron, when roughly 70,000 a year would have been needed for genuine reindustrialisation.

The EY European Attractiveness Survey points the same way. Actual FDI projects, meaning foreign direct investment in real facilities built by foreign companies, fell in France from 1,194 in 2023 to 852 in 2025. That is a 29% drop in two years.

France still ranks first in Europe. But the pipeline is thinner, even as the announcements grow larger.

SoftBank’s €75 billion: read the small print

SoftBank’s "up to" €75 billion is a ceiling, not a commitment. Phase 1, worth €45 billion for 3.1 GW of data centre capacity by 2031, has named sites and partners, including EDF, the state-owned energy utility. Phase 2, another €30 billion, has no timeline and no disclosed financing plan.

That matters because SoftBank is not sitting on a fortress balance sheet. S&P rates the company BB+ with a negative outlook, below investment grade. It carries roughly $123 billion in net debt and faces a $50 billion refinancing wall by the end of 2026.

Its balance sheet leans heavily on paper gains from a $64.6 billion stake in OpenAI, a privately held company whose shares cannot be easily sold. Analysts describe this as the "old playbook". In 2016, SoftBank chief executive Masayoshi Son pledged $50 billion to the United States alongside Donald Trump. Much of that money ended up in WeWork, which collapsed.

The same balance sheet is now being used to promise $100 billion in the US and €75 billion in France.

Who gains, who pays

France does have a real advantage in this race: nuclear power. Its 57 reactors produce cheap electricity around the clock, at roughly 73 USD/MWh wholesale, compared with 100 USD/MWh in Germany. For AI data centres, which need constant power, that is more useful than Spain’s cheaper but intermittent solar.

France now offers large industrial users long-term contracts linked to EDF’s production cost rather than wholesale market prices. Bruegel, the Brussels-based think tank, has described such arrangements as Europe’s "under-the-radar industrial policy".

The trade-off is jobs. Data centres create very few permanent roles once construction is finished. Brookings Institution research shows a typical large facility employs 50 to 400 people after it opens. Building them is labour-intensive. Running them is not. For the sums SoftBank is discussing, a traditional factory would leave behind far more lasting employment.

The power demand is not marginal either. SoftBank’s full plan would consume roughly 45–50 TWh a year, about 10% of France’s total electricity output. Whether the northern transmission grid can absorb 3.1 GW of new load by 2031 remains unanswered. Neither RTE, the grid operator, nor EDF has publicly addressed it.

What remains uncertain

Some Choose France projects do become real. Orano-XTC broke ground on a battery materials plant at Dunkirk in May 2026, three years after its announcement at the summit.

Others slip away more quietly. France’s biggest European battery venture, ACC, had to bring in Chinese technicians to keep its own factory running.

The issue for Europe is whether this model, built on large pledges, generous energy pricing, limited job creation and no public accounting, amounts to an industrial strategy or a communications strategy. Until France publishes conversion rates, everyone else is left counting press releases.