ZACH Multimedia

23/10/2026

India mounts costly defence of rupee

Central bank has spent about $90bn of money raised from diaspora to shore up currency

[Image - The rupee is one of Asia’s worst-performing currencies this year, having fallen 5% against the dollar © Amit Dave / Reuters]

By Chris Kay in Mumbai, Financial Times, 22 Oct 2026

India is mounting an increasingly costly defence of the rupee, deploying tens of billions of dollars and resorting to extraordinary market interventions to protect a currency that has come under pressure from elevated oil prices and investor outflows.

The central bank since June has already used about $90bn of the more than $140bn it raised from India’s diaspora to prop up the rupee, including through sales to the spot market and currency swaps to absorb surplus liquidity, according to two foreign-exchange bankers with knowledge of the moves.

“They have been aggressive,” said one of bankers. The Reserve Bank of India did not respond to a request for comment.

The rupee is one of Asia’s worst-performing currencies this year, having fallen 5 per cent against the dollar. Rising crude prices from the war in Iran have driven up the cost of energy imports for one of the world’s largest oil consumers, while record foreign selling of Indian equities has exacerbated the rupee’s weakness.

As the currency neared new lows this month, the RBI took unexpected steps to curb volatility.

They included opening a special window to absorb state oil companies’ dollar purchases away from the spot market; cutting the threshold for rupee-linked derivative transactions without proof of underlying exposure; and requiring dealers to maintain a 20 per cent cash reserve on forward contracts with the central bank.

Those moves, which came on top of earlier measures such as curbing short positions against the rupee and courting deposits from Indians living abroad, have helped bolster the currency.

The rupee has strengthened 0.7 per cent in the past week to trade at about Rs94.5 a dollar on Thursday, a six-week high. Foreign inflows to take part in the public offering of telco Jio Platforms, set to be India’s biggest-ever listing, have also helped.

But bankers and economists said the RBI could not keep spending down its foreign exchange reserves at its current pace. The central bank’s reserves have fallen by about $60bn from an early September peak of $785bn. Its net forward dollar liabilities — dollars which it will have to pay in the future — ballooned to $200bn in August from about $137bn in July, according to official data.

“It is not sustainable,” said Dhiraj Nim, a Mumbai-based economist and foreign exchange strategist at ANZ Group. “FX reserves are finite and the forward book obligations are large. The RBI will not draw down reserves to an extent that reserve adequacy becomes a concern.”

Nim said the central bank was also reluctant to raise interest rates aggressively to shore up the currency because of the potential hit to economic growth. India’s year-on-year GDP growth slowed to 7.8 per cent in the quarter to the end of June from 8.6 per cent in the previous quarter.

“Policy rate hikes will impact the economy more widely, explaining the RBI’s reluctance to hike aggressively,” he said.

This month, the central bank raised interest rates for the first time since 2023, by a quarter point to 5.5 per cent. The move came after India’s inflation accelerated to 5.5 per cent in September, nearing the RBI’s tolerance limit of 6 per cent.

“If the currency becomes disorderly, more aggressive hikes cannot be ruled out,” said Nim, adding that the RBI seemed to have “drawn a line in the sand” at Rs97 per dollar.

In an interview with local media this month, RBI governor Sanjay Malhotra reiterated the central bank’s commitment to a stable rupee and said it did not target a certain exchange rate or band.

[‘Cash is making a comeback’ as India raises fees on digital payments]

He also rejected suggestions that this month’s measures amounted to capital controls, arguing they were meant to ensure currency markets catered to genuine hedging needs rather than speculative bets.

Krishna Srinivasan, Asia-Pacific director at the IMF, said in a media briefing in Mumbai on Wednesday that India should allow the exchange rate to move as the first line of defence against the ongoing oil shock.

“Malhotra’s been more open and explicit about letting the market broadly determine where the exchange rate should be, with the bank’s role simply to curb any excess volatility,” said Miguel Chanco, chief emerging Asia economist at Pantheon Macroeconomics.

“This ethos is certainly now being tested by the fallout from the war in the Middle East, which is Malhotra’s first real ‘crisis situation’.”

Central bank has spent about $90bn of money raised from diaspora to shore up currency

22/10/2026

Soaring AI Bills Are Vexing Wall Street Executives

The trajectory of computing costs is alarming finance chiefs already struggling to put a price tag on AI.

[Image - Global banks are wrestling with surging AI bills as executives deploy the new technology across workflows. - Illustration by Hunter French]

By Rthvika Suvarna, Bloomberg, 22 October 2026

— With assistance from Nil Codina Martinez and Paige Smith

Takeaways by Bloomberg AI

> Chief financial officers are looking for ways to control costs associated with AI adoption, as surprise bills have become an issue.

> Agentic AI, which can tackle complex tasks autonomously, is emerging as a primary driver of rising costs due to its unpredictable computing power consumption.

> Banks are exploring strategies to manage AI costs, including restricting access to powerful models, testing smaller models, and considering alternative providers, such as Chinese models.

No one ever wants to give a technology team a blank check. Yet for many chief financial officers, that’s exactly what they were asked to do in the early rush to embrace AI. Now, after one-too-many surprise bills, finding a way to control costs has become an imperative.

It’s a simmering issue in banks, where technology budgets are already under pressure from competing demands ranging from legacy IT modernization to cybersecurity. Some of the greatest hopes for productivity improvements are pinned on agentic AI, which can tackle complex tasks autonomously, but are inherently difficult to budget for as the amount of computing power each task consumes can be unpredictable.

“I actually don’t think anybody in the organization has a good handle on it — if I am going to do XYZ, what is going to be my cost,” said Bhavi Mehta, a partner at Bain & Co. and its global lead for advanced analytics in financial services. “As AI moves toward more sophisticated reasoning and agentic workflows, the tokens required for a given task can vary significantly. That makes understanding cost alongside business value increasingly important as banks scale AI,” he added.

Only 12 of the 50 largest global banks tracked by AI benchmarking firm Evident have publicly attributed projected or realized returns to their adoption of AI, according to its latest report. Yet it’s a rare CEO who doesn’t talk up the potential benefits. Rising investment in the technology came up during third-quarter earnings calls at several US banks from Citigroup Inc. to Goldman Sachs Group Inc., but only a few executives had precise figures for their AI expenditures or the financial benefits they are generating.

So as the industry moves from the experimentation to the deployment stage, Bloomberg News spoke with more than a dozen engineers, executives and advisers about their experiences and strategies.

Match the Model to the Job

While tokenmaxxing — the practice of engineers using AI as much as possible — has become notorious — that’s not always the cause of token usage blowout.

When Deutsche Bank AG’s monitoring flagged individual employees consuming unusually large volumes of tokens, many turned out to be doing useful work, said Christoph Rabenseifner, the bank’s chief strategy officer for technology, data and innovation. Instead of cutting them off, the bank steered that usage in a more productive direction. It also tested one task on a smaller, specially trained model instead of a large frontier model, cutting costs several times over, Rabenseifner added.

[Image - Christoph Rabenseifner - Source: Deutsche Bank AG]

Banco Santander SA runs large frontier models only to check whether an application meets its quality bar, then shifts proven workloads to lighter, cheaper ones. The approach helped cut ex*****on costs from about €5 ($5.61) to €0.03 per loan approval process, said Ignacio Bernal, the bank’s chief artificial intelligence officer.

“When you move AI from pilots to production across a global bank, inference cost stops being a rounding error and becomes a real operational cost line,” Bernal said, noting some processes run millions of times a year, driving up computing expenses significantly.

Canadian Imperial Bank of Commerce has found that its most powerful models, including Anthropic PBC's Opus, can cost C$3,000 ($2,130) per employee a month, a bill that could add up quickly across thousands of workers. As a cost-control measure, the bank restricted access to a few hundred employees, while capping its latest in-built workspace CAI 2.0, still in testing, at 250 users who can each spend between C$4 and C$250 a month.

“It’s too powerful to have 50,000 people just using it without control,” Jaime Tatis, CIBC Chief Data and Artificial Intelligence Officer, said in a July interview.

[Image - Key Speakers At The All In Event - Jaime Tatis - Photographer: Graham Hughes / Bloomberg]

Agentic AI is emerging as a primary driver of rising costs. Big banks announced 19 agentic AI use cases in the first quarter of 2026, up from eight the previous quarter, according to Evident. In the second quarter, the number halved – not because banks are cooling on agents, but because the firms are asking them to take on more complex work. Wells Fargo & Co., for instance, wants agentic AI to master 1,100 steps in the mortgage process before deploying it across workflows, the Evident report said.

Unlike conventional AI tools, agentic systems autonomously execute complex, multistep tasks, which include analyzing financial statements, tracking borrowers, verifying documents and detecting money laundering activity. This processing demands significant computational resources, translating into heavier token consumption, and in turn, drives up monthly costs.

At one Asian lender after some teams burned through their allocation of tokens in the first week, the bank decided to pool tokens and build an agent that trims prompts before they reach a model, according to a senior data executive who asked not to be identified talking about internal matters.

Token Use For Common AI Tasks Varies Widely
This makes bills hard to forecast for banks
Source: Data compiled by Wolfe Research

Consider Alternatives

The cost question has brought Chinese models into the conversation. A July JPMorgan Private Bank note said a GLM model from Beijing’s Z.AI performed on par with the best model available in March at more than 80% less than the priciest frontier model at the time.

Singapore’s biggest lenders are experimenting with a variety of models. United Overseas Bank Ltd.’s neutral AI framework is open to Chinese providers, and DBS Group Holdings Ltd. says it is not overly reliant on any one provider, whether it is US or Chinese ones.

Regulators have flagged some concerns. Canada’s banking watchdog recently asked staff to draft minimum standards for how lenders use frontier AI models, including on security, while the US Federal Reserve Vice Chair for Supervision Michelle Bowman has warned banks to shore up their systems against AI-driven threats.

That makes less-vetted models, particularly foreign ones, harder to justify to boards. Risk chiefs also worry that AI agents, once deployed, can behave unpredictably inside a bank's own systems, a concern that has gained traction in recent months.

National Australia Bank Ltd. cited security and governance concerns when considering alternatives to US-built AI models, including Chinese ones, Bloomberg News reported in August.

Chinese AI Models Undercut US Rivals on Cost
Cost per 1 million output tokens for selected large language models, in US dollars
Note: Prices are standard output-token rates listed on OpenRouter as of Oct. 21, 2026. Price may vary by inference provider, routing option and usage tier.
Source: Data compiled by OpenRouter

Oversight Strategy

During the early rush to cloud services more than a decade ago, banks took at least five years to control costs that rose with usage and bring them under a dedicated budget line, according to J.R. Storment, executive director of the Tokenomics Foundation. What’s different now is that financial chiefs are reacting to AI cost spikes within 6 to 18 months of adoption, making it difficult to put a final price tag on premium AI models, he added.

Nitin Madan, formerly CFO at Barclays Plc’s India operations, remembers the pressure well. When he asked his colleagues on the technology side which AI models they planned to deploy and what they would cost, he was told such questions were “making everybody uncomfortable,” Madan said.

As new AI tools and upgrades hit the market in rapid succession, Madan said AI spending across the industry can’t be spread over several years the way most software costs are, so it hits profits right away. No one can say with confidence whether today’s models will still be in use three years from now, he added.

At JPMorgan Chase & Co., some employees were “spending more on tokens than their salary” and yet the company wasn’t rationing AI usage at that time, Zachery Anderson, chief data and analytics officer for the bank’s payments division, told Semafor in June. JPMorgan has since set monthly spending limits of $2,000 on Anthropic’s Claude Code for certain engineers, Business Insider reported this month. The bank’s CFO Jeremy Barnum said on the lender’s earnings call this month that its annual AI expense was about $700 million, though a precise figure is hard to pin down because the technology is now spread across the entire bank.

[Image - JP Morgan Chase & Co. Ahead Of Earnings. The JPMorgan Chase & Co. headquarters in New York. - Photographer: Gabby Jones / Bloomberg]

At many other banks, leadership has also shifted from treating AI spending as virtually limitless to finding ways to contain it.

Alex, a senior machine-learning engineer, at a large European bank was asked to cap his team’s AI spending in February this year. The trigger was an €80,000 bill the team had piled up after building an advanced chatbot and deploying it across thousands of workflows. The sum, though hardly outrageous for a large bank, signaled to the finance chief that it might snowball into millions if left unchecked. For Alex, who asked not to be identified by his full name to avoid professional repercussions, the question quickly shifted from whether the chatbot worked to whether the bank could afford to run it at scale.

To cut costs, he reworked the chatbot so it cached answers to frequently asked questions and resurfaced them whenever a user asked again. Simpler requests went to open-source models. Only token-intensive prompts, the ones requiring deeper searches, went to premium models from platforms like ChatGPT or Claude. That brought the AI bill down by roughly a third.

Some banks are tackling the forecasting problem earlier. French lender BNP Paribas SA approves every AI project after its running cost is calculated. The group’s internal platform sets a token price, multiplies it by expected volumes and weighs the total against the value the tool should deliver, said Hugues Even, the bank’s group chief data officer. BNP covers most of its computing needs with its own chips, bought as multiyear investments, and supplements them with cloud providers.

BNP spreads oversight of AI across committees chaired by the heads of technology, risk, finance and human resources, plus a group AI committee led by the chief executive officer. “Each one is taking his share of the cake,” Even said. The finance chief controls the cost of deploying AI across the group, he added.

[Image - French Banks Ahead of Earnings. The BNP Paribas headquarters in Paris. - Photographer: Nathan Laine / Bloomberg]

Accountability Focus

Many lenders are contending with the speed of change, as newer and pricier models arrive before the last integration is complete. And to help them measure and rein in the potential cost surges, a cottage industry is emerging.

Ramp, which helps companies track and control token spending, says one in three businesses using its tool found out where they could shift from costlier frontier models to cheaper ones without compromising quality.

Rogo Technologies Inc., an AI startup that works with bankers, says it pools demand across clients to negotiate discounts with AI labs and lets banks set token budgets down to individual users.

Tokenomics Foundation, whose members include JPMorgan, Bank of New York Mellon Corp., KPMG LLP, and Google Cloud, has been created to set shared industry standards for tracking AI costs, which are often opaque. It seeks to measure what companies get back in return.

Access has become a hiring issue, too. At the Asian lender, developer candidates ask in job interviews about their token budget alongside their salary. Engineers ask which models they will get and what limits apply, Madan the former CFO said.

David Hardoon, who led AI enablement at Standard Chartered Plc and is now head of advanced AI for Southeast Asia at Accenture Plc, argues every large organization needs a chief AI officer with real authority.

While several banks already have one, few have given the role enough power to impose discipline on forecasting and deployment without slowing innovation. A chief AI officer could own that question but only with the authority to act across the whole organization, Hardoon said.

For now, keeping a tab on the cost equation during the AI gold rush can cause internal tensions, especially when bank executives feel they cannot risk falling behind on adoption.

The thinking is: “If we’re not doing it, our competitors are doing it,” said Vitaly Gordon, CEO of Faros AI, a software firm that helps banks and tech companies tie AI spending to results.

This arms race mentality across the industry is translating into cash burn and causing a situation where he said “in most cases, it’s just a CFO getting a bill and screaming.”

https://www.bloomberg.com/news/articles/2026-10-22/how-wall-street-banks-are-grappling-with-soaring-ai-bills?srnd=homepage-middle-east

The trajectory of computing costs is alarming finance chiefs already struggling to put a price tag on AI.

21/10/2026

What Better Tribute To Maa Durga On This Vijayadashami Day, Than This Raag Devoted To Maa Durga, And Magically Crafted Through The Shehnai By One Of India's Greatest Maestros, Ustad Bismillah Khan, Among The First To Elevate The Status Of That Reeded Woodwind Instrument To Its World Acclaimed Levels Now. 🙏🙏🙏

[Durga is a raga in Hindustani Classical music. It is called Shuddha Saveri in the south. The raga is to be sung during late night.
Bismillah Khan, referred by honorific title 'Ustad' was an Indian Musician. He was the first one to elevate the status of the instrument and created immense importance. He was awarded with Bharat Ratna in 2001. He is popularly known for his vision and aim of spreading peace and love by the medium of music. He passed away on 21st August 2006 because of a Cardiac attack.]

Durga is a raga in Hindustani Classical music. It is called Shuddha...

21/10/2026

ZACHMultimedia Wishes A Blessed Vijayadashami

To All 8 Billion Plus Fellow Travellers

20/10/2026

India’s democracy once went looking for voters. Not any more

The franchise is being recast as a privilege citizens must repeatedly prove they deserve

[Image - © Ewan White]

By Milan Vaishnav, Financial Times, 19 Oct 2026

[The author is a senior fellow at the Carnegie Endowment for International Peace and is writing a book on India’s electoral democracy]

A democracy that once went looking for voters is increasingly demanding that voters prove they belong. That reversal is at the heart of a growing political confrontation in India.

In recent days, opposition parties and civil society activists have taken to New Delhi’s streets demanding the resignation of the country’s chief election commissioner. The mobilisation has been catalysed by the Cockroach Janta Party, an upstart Gen Z movement whose earlier protests over exam fraud forced the resignation of India’s higher education minister.

At the heart of the current protests lies the Special Intensive Revision (SIR), a nationwide overhaul of voter rolls launched last year by the Election Commission of India, ostensibly to remove ineligible voters, including alleged non-citizens often labelled “infiltrators” by the government. To date, more than 130mn names have been removed from draft voter rolls.

Many reflect deaths, migration or duplicate registrations, while others have been reversed. Critics say the exercise is opaque and especially burdensome for disadvantaged voters who may lack basic documentation. How many eligible voters have been wrongfully excluded is unclear.

After gaining independence from the British in 1947, India distinguished itself by extending the franchise to all adult citizens regardless of wealth, literacy or social standing. But its founders understood that granting the vote meant little without ensuring citizens could exercise it. Indeed, India began preparing its first electoral rolls before its 1950 constitution came into force.

The constitution established a powerful, centralised Election Commission that assumed responsibility for enrolling every eligible voter, regularly conducting house-to-house surveys spanning glaciers, deserts and jungles. In contrast to many established democracies, where registration has historically been dependent on individual initiative, in India the state assumed the burden.

Maintaining accurate electoral rolls in any democracy is essential. Election authorities carry a dual responsibility: they must simultaneously prevent ineligible individuals from voting while ensuring eligible citizens are not excluded. When these come into tension, India historically prioritised widening the electorate.

Under the SIR, even Indians who have voted in numerous past elections are being asked to prove their citizenship. This is a striking reversal of India’s founding electoral bargain: the state is increasingly shifting the burden of establishing eligibility to voters. The demands are especially problematic because India has no universal document establishing citizenship. A tiny share of Indians holds passports, and while nearly all possess Aadhaar, a unique biometric ID number, the government recognises it as proof of identity, not citizenship. Predictably, the challenge of proving citizenship falls disproportionately on poor people, women, migrants and minorities.

The consequences of the SIR are not hypothetical. In West Bengal’s April 2026 state election, more than 9mn names were removed from voter rolls. While 6.4mn were initially classified as dead, absent, relocated or duplicates, another 2.7mn were removed after software flagged discrepancies in their records. Of this group, over 80 per cent appealed, but just 1,607 people were reinstated in time to vote. Months later, most appeals are unresolved. Yet 93 per cent of decided appeals resulted in reinstatement, raising serious questions about the accuracy of the original deletions.

The SIR also threatens one of India’s greatest institutional assets: an Election Commission whose sweeping powers have rested on its reputation for impartiality. Allegations of partisan bias and deference to the Bharatiya Janata Party government led by Prime Minister Narendra Modi, together with reports of dissent within the commission, have eroded confidence in its independence.

India’s ability to conduct elections remains impressive; more than 640mn people voted in its last general election. But disputes over voter eligibility, bureaucratic obstacles and allegations of partisan manipulation are reminders that administrative sophistication alone cannot guarantee electoral integrity. Amid the uproar over the SIR, the Election Commission has belatedly instructed officials to identify and re-enrol eligible voters omitted from the list.

But even if wrongly deleted voters are eventually reinstated, they can never recover the votes they were denied. A country that once went to extraordinary lengths to ensure all citizens could vote is now busy placing new obstacles in their way.

The franchise is being recast as a privilege citizens must repeatedly prove they deserve

19/10/2026

America’s $10tn AI infrastructure boom is starting to fray

A vast bet on the future of the technology is running into the unforgiving economics of the physical world

[Image - © Joe Lamberti/Bloomberg]

By George Hammond and Ryan McMorrow in San Francisco, Kieran Smith in London and Eva Xiao in New York, Financial Times, 18 Oct 2026

Additional reporting by Tim Bradshaw and Michelle Chan. Data visualisation by Clara Murray

Cracks are appearing in America’s debt-laden $10tn AI infrastructure boom as project delays, public backlash and shaky investor confidence bring a reckoning for the data centre build-out underpinning the US economy.

A rush for computing power in recent years created what multiple people involved characterised as a “Wild West” market. They said money has flooded into brand-new companies and due diligence for new projects has been expedited, with AI labs such as Anthropic and OpenAI accepting higher short-term costs to move as fast as possible.

More than three years into the AI boom, these ambitious data centre plans are running into reality. At least 47 large projects have been delayed, withdrawn or blocked in 2026, according to data shared by RegLayer, a regulatory intelligence company, which described it as a “conservative estimate” focused on the largest sites.

Analysts at Bernstein this month estimated only about 35 per cent of the planned US data centre pipeline was credible. More than 1,000 communities across the US have tried to introduce moratoria or other restrictions on AI infrastructure this year, according to FT research.

AI labs and Big Tech groups have committed to spend trillions of dollars over the coming years on computing capacity, or “compute”.

But those massive rewards can only be realised if a group of recently launched “neocloud” data centre providers first manage to finance the unprecedented construction effort.

Higher interest rates and a souring bond market are now threatening their ambitions, and with them the AI build-out, said a number of people involved in the market.

This week investors began repricing risks in the sector: debt tied to many AI infrastructure groups, including upstart cloud provider CoreWeave, sold off sharply and share prices for listed companies slumped. CoreWeave’s stock lost almost 25 per cent during the week.

Higher borrowing costs and rising inflation have added to mounting questions about whether the revenues of leading labs such as Anthropic and OpenAI will be sufficient to cover the costs.

Plans by a procession of new developers and cloud providers including SoftBank’s SB Energy and Nvidia-backed Firmus to go public have been delayed or mothballed, with OpenAI’s decision to delay its own initial public offering further damping sentiment.

[Image - A data centre site under construction in Saline Township, Michigan. © Bloomberg]

Construction delays also raise the prospect that tenants, typically Big Tech companies or AI labs, could invoke clauses letting them walk away if projects are not delivered on time.

High-profile projects have faced lengthening timelines, with setbacks affecting several Oracle sites being built for OpenAI’s Stargate data centre initiative.

Former crypto miners such as TeraWulf and Cipher have also renegotiated leases and slightly lengthened delivery schedules on projects for Anthropic in recent months, in part because of design changes from the AI lab and its partner Fluidstack.

The adjustments can be costly. When amending its lease with Fluidstack at a Texas site last month, Cipher secured more time to deliver and agreed to absorb the first $359mn of cost overruns. The changes required roughly $750mn of additional equity, partly funded this week by an $850mn convertible bond.

Anthropic can also walk away from its $45bn deal with Nscale, a London-headquartered neocloud launched in 2024, if the group does not hit financing or delivery timelines for a major project in West Virginia.

The Nvidia-backed neocloud is targeting a New York IPO at a $35bn valuation, but bankers say that will be difficult to pull off in the near term given volatile market conditions. A person familiar with the matter said Nscale hoped to list this year. Nscale declined to comment.

Nscale had just 25,000 advanced AI chips online in total as of August, compared with the nearly 500,000 it plans to operate.

Jensen Huang, chief executive of Nvidia, was more upbeat on Wednesday in London, telling a crowd that his company had helped build up Nscale in a year’s time.

“Last year you were zero,” he said, pointing to an Nscale executive. “Look at them now: a proud English start-up, it could be the most valuable UK company created in the last 10 years.”

Confidence in the data centre boom has been underpinned by the growth of OpenAI, Anthropic, Meta, Google and other AI groups.

OpenAI chief executive Sam Altman has staked his company’s future on the effort, claiming “compute is destiny”.

A crop of businesses offering to source land, semiconductors, power and data centres has sprung up in response, capitalising on rampant demand and generous equity and debt markets to scale fast.

AI labs such as Anthropic and OpenAI have reasoned they are better off risking the loss of a portion of their investment by moving fast — even where they rely on inexperienced partners — than suffering a shortfall of computing power that prevents them delivering their models, according to people with knowledge of the matter.

They have also structured contracts to protect themselves should partners fail to deliver. An OpenAI spokesperson said the company’s approach balanced “speed, cost and reliability.”

The resulting outlay on AI infrastructure is unprecedented: US spending is projected to surpass $10tn between 2025 and 2032, according to estimates by Stijn van Nieuwerburgh, an economist at Columbia Business School, in a paper for the Brookings Institution.

That equates to about 3.6 per cent of US GDP a year, a far higher proportion than previous infrastructure waves including railways, highways and electrification.

But investors this week have increased their scrutiny of the immense capital needs of America’s AI build-out.

CoreWeave, a small crypto miner four years ago, was among the first to capitalise on the AI boom but has been at the centre of a sell-off as investors fret over its ability to keep financing the billions it will spend on Nvidia hardware and data centre leases.

One of its unsecured bonds sank into distressed territory this week, while bonds funding data centres leased to CoreWeave also slid. A $1.1bn bond for a site in Virginia has lost about 5 per cent of its value in the three weeks since it priced.

[Image - A data centre being built in Saline Michigan for Oracle and OpenAI © Jim West / UCG / Getty Images]

SB Energy, majority owned by Masayoshi Son’s SoftBank and backed by OpenAI and Nvidia, has not completed a single data centre, and has $174bn in capital needs in order to satisfy contracts worth $430bn in future revenue, largely from OpenAI. The group has delayed plans for a public listing at a $50bn valuation.

Arkady Volozh, chief executive of another Nvidia-backed neocloud Nebius, told a Citi conference in September that building out its planned 5 gigawatts of capacity would take about $250bn of financing. “So far, we financed maybe 20 per cent of that,” he said.

These financial challenges come on top of the inevitable hurdles facing any large infrastructure project as well as growing public discontent with data centre construction.

OpenAI’s $500bn Stargate infrastructure scheme, which sought to bind Altman’s ambitions in the sector to President Donald Trump’s, has run into a litany of issues including thwarted planning permissions, project delays and the departure of senior executives — including the group’s computing lead Sachin Katti this week.

Securing building permissions has become increasingly difficult across the sector while utility regulators have been refusing grid connections unless projects can supply their own power, according to data centre executives.

“People aren’t just dropping their shorts for data centres anymore,” one executive said. “It’s much harder today to finance a data centre.”

A vast bet on the future of the technology is running into the unforgiving economics of the physical world