For the ongoing wave of artificial intelligence investment in the US to yield a financial return, businesses and consumers would need to spend approximately 9% of national income on the technology, according to calculations.

This proportion is equivalent to the entire budget the American public allocates to food, double all energy expenditures, or roughly seven times the total amount paid for telephone, internet, and digital streaming services.

A study prepared by Columbia University finance professor Stijn Van Nieuwerburgh and presented at the Brookings Institution, reported by The Wall Street Journal (WSJ), revealed that artificial intelligence infrastructure construction has become the largest investment drive in US history.

Based on current data centre construction plans, average costs, potential project cancellations, and an expected 10% unleveraged return on capital, calculations show that the sector must generate $3.5 trillion in revenue by 2032.

Under the assumption of 4% annual nominal growth, this figure corresponds to 8.8% of US gross domestic product (GDP).

Capacity expansion may lower prices

The foundation of current revenue projections rests on the assumption that artificial intelligence companies such as OpenAI and Anthropic will be able to sustain existing processing power prices into the future.

Stating that the current deficit in processing capacity is keeping profit margins high, Van Nieuwerburgh noted that companies will not be able to sustain today’s scarcity pricing in an intense competitive environment by 2032, when capacity is projected to quadruple.

Sharp price declines seen in past technology investments show that comparable crises have occurred before.

As a result of laying new lines and expanding capacity, bandwidth prices on transatlantic fibre-optic links between London and New York plunged by 96% between 1997 and 2001, a process that triggered the bankruptcy of numerous telecommunications companies.

Could falling prices boost demand?

While the processing capabilities of artificial intelligence models double every four to five months, data from Epoch AI show that the cost for a given capability level has been falling by 47% per quarter since 2023.

Optimists, drawing on the concept known in economic literature as Jevons’ Paradox, argue that expanding capacity and lower costs will multiply demand, thereby driving corporate revenues higher.

The $300 billion combined annual revenue estimate set for 2030 for OpenAI and Anthropic by an expert panel led by Ezra Karger of the Federal Reserve Bank of Chicago has already been surpassed.

The annualised combined revenue run rate of the two companies has reached approximately $180 billion and continues to post double-digit growth each quarter.

Against advocates who view the technology as a third factor of production alongside capital and labour, forecasts indicate that the growth rate will decelerate and the law of diminishing returns will take effect as the system becomes widespread.

Productivity gains do not translate directly into revenue

Following the price drops and productivity surge seen from 1966 onwards during the personal computer revolution, corporate computer expenditures flattened out at around 0.8% of GDP by approximately 1984.

A recent study conducted by Harvard University doctoral students Fiona Chen and James Stratton, examining software projects on the Jellyfish platform, pointed to a similar structural ceiling.

The researchers determined that artificial intelligence assistants increased the number of lines of code written by 12%, while autonomous software agents increased it by 30%.

Conversely, because users spent more time reviewing, interpreting, and modifying code, the ultimate contribution to completed projects remained low and was found to be “statistically insignificant”.

Experts emphasise that even if artificial intelligence boosts productivity, this outcome may not translate directly into the massive profit and revenue figures anticipated by investors.

The world’s richest man, Elon Musk, has assumed the co-directorship of a Pentagon initiative focused on the future of warfare, known as “Project Meridian”.

Musk’s new role was announced by US Secretary of Defence Pete Hegseth.

Musk, who has long expressed his conviction that wars will ultimately be fought with autonomous unmanned aerial vehicles, will advise the project as co-director alongside Palmer Luckey, founder of defence start-up Anduril, and former Speaker of the House of Representatives Newt Gingrich.

In a memorandum issued at the Pentagon, Hegseth stated that the group would “examine the battlefields of the future” and “determine which weapons and technologies warfighters must employ to achieve dominance in these environments.”

During his “State of the Force” address at Marine Corps Base Quantico, Hegseth said:

“The best predictors of future conflict do not reside exclusively within the Pentagon. Obvious biases and risks arise when we task ourselves with both framing the questions and answering them.”

Hegseth stated that this initiative would commence immediately and that, following his address, he would convene with Musk, Luckey, and Gingrich at a secure location.

Project Meridian will have 120 days to “ruthlessly map the trajectory of wars, domains, and technologies”, a process that will culminate in the public disclosure of its findings alongside a classified annex.

Hegseth outlined an expansive mandate extending “from beneath the surface of the Earth to beyond the Moon.”

Rather than formulating new military strategies or policies, the panel will seek to identify “the domains we must seize and the capabilities we must master”, focusing on the effort to “discover, develop, and field” the weapons and systems that next-generation American troops may require.

The group is expected to submit a report containing recommendations to him by the end of January.

In 2024, Musk remarked: “Future wars will be entirely about drones and hypersonic missiles.” This was merely one of several similar statements he has made in recent years.

For Musk, whose oversight role at the Department of Government Efficiency (DOGE) ended in turmoil and escalated into a dispute with President Donald Trump over Trump’s spending bill, this appointment marks his formal return to government in an official capacity.

Musk and Trump ultimately reconciled, and Musk attended a meeting on artificial intelligence safety at the White House this week alongside other technology leaders.

Meridian forms part of a broader push announced by Hegseth to restructure the military around autonomous warfare and rapidly advancing technologies.

Hegseth announced the establishment of the Autonomous Warfare Command (AUTOWARCOM), a new four-star combatant command endowed with what he termed “service-like authorities” to scale autonomous and robotic capabilities across the joint force.

The Department of War will also begin phasing in new occupational frameworks across all military branches to establish specialised career tracks for what Hegseth described as “the next generation of autonomous warfighters.”

“We should have conceived an Autonomous Warfare Command a decade ago,” Hegseth said, explaining that Meridian aims to gaze far enough ahead to enable the military to anticipate the next technological shift rather than lag behind.

A cyberattack targeting the US Department of War’s personnel database has resulted in the leak of personal information belonging to approximately 3 million people.

Speaking to ABC News, a Pentagon official stated that the system accessed by unauthorised individuals contained the records of 2,760,000 living persons and 294,000 deceased individuals.

The Military Times portal, which first broke the news, had reported the number of affected individuals as approximately 4 million based on two sources. The Pentagon official subsequently conveyed different figures to ABC News.

The leak encompasses Social Security numbers and duty information belonging to military personnel and civilian employees. According to an official notification examined by Military Times, the compromised records may also include names, dates of birth, contact information, sex, race, and military occupational specialties.

The unauthorised access to the information system of the Defense Manpower Data Center (DMDC) lasted for approximately nine months, between October 2025 and 16 July 2026.

ABC News reported that the access in question was obtained by a small number of third-party users. The vulnerability was closed after it was identified.

The DMDC is considered one of the Pentagon’s primary personnel records centres. More than 60 million records belonging to active-duty personnel, reservists, civilian staff, contractors, retirees, veterans, and military family members are stored at the centre.

The Pentagon has not detected any evidence that the leaked data has been misused. Military Times reported that affected individuals were offered identity restoration and credit history monitoring services.

A similar data breach previously occurred on the Federal Bureau of Investigation’s (FBI) recruitment website, FBIJobs.gov. According to information obtained by ABC News from internal communications and sources, the FBI is considering the possibility that data belonging to its entire staff may have been stolen.

The New York Times (NYT) examined a portion of the stolen FBI records. Home addresses, telephone numbers, official email addresses, Social Security numbers, dates of birth, hiring dates, and emergency contact details for relatives were identified within these documents.

The database also contained unit designations, duty roles, and information regarding the supervisors of personnel. Some records revealed assignments within counterintelligence and counternarcotics units, as well as departments examining threats originating from Russia, China, and Iran.

Ciaran Martin, the former head of the UK National Cyber Security Centre, noted that this type of breach could directly affect the FBI’s operational capabilities.

The hacker group known as ShinyHunters had announced that it had seized medical data and security clearance records alongside files belonging to tens of thousands of active and former FBI employees.

Experts evaluating the matter for the NYT warned that this information could be used to track agents, threaten their families, or compile dossiers by foreign intelligence services.

The ShinyHunters group initially threatened to release the data unless the bureau withdrew an advisory it had published concerning the group’s attack methods.

The group later asserted that it had never intended to leak the information and characterised its action as an advertising campaign.

In a report published in May, Reuters noted that the personal data of US military personnel had been used in surveillance and attack preparations.

According to the agency, Washington’s adversaries gained the ability to pinpoint areas where troops were concentrated by exploiting commercially available location data. US lawmakers at the time criticised the Pentagon for failing to adequately protect the personal data of military personnel.