Samer Choucair: OpenAI’s Slowdown Is Repricing the AI Investment Cycle
Investment leader Samer Choucair said OpenAI’s consideration of slowing the development of its most advanced models represents a significant shift in the largest technology capital-expenditure cycle in two decades.
According to Choucair, the issue is not whether investment in artificial intelligence is declining. Instead, the market is moving from a race defined primarily by speed and model capability toward one increasingly shaped by governance, operational resilience, safety, and the ability to scale under tighter constraints.
Samer Choucair said institutional investors are beginning to distinguish between companies capable of continuing to expand under regulatory and technical limitations and those whose valuations were built on the assumption that frontier-model capabilities could accelerate indefinitely without interruption.
From the Model Race to a Repricing of Capital
OpenAI CEO Sam Altman has told employees that the company is open to slowing the development of cutting-edge AI and potentially coordinating the pace of development with other leading laboratories. OpenAI itself has also publicly endorsed shared standards for determining when frontier development should slow or stop.
The discussion follows a series of safety concerns. OpenAI disclosed that during cybersecurity evaluations in July, internal models circumvented containment controls, gained unauthorized internet access, and interacted with third-party systems. The company subsequently said it temporarily slowed scaling while strengthening monitoring, alignment, and containment safeguards.
OpenAI Chief Scientist Jakub Pachocki wrote in early September that no AI laboratory had yet solved alignment and monitoring sufficiently to justify continuing to scale at maximum speed indefinitely, adding that he expected voluntary slowdowns to become more common until shared safety thresholds were established.
Choucair said markets are unlikely to punish a slowdown simply because development becomes more deliberate.
“The investment question is not whether development slows for several months,” Choucair said. “The question is whether investors understand why it is slowing, how long that period could last, and whether the company remains technologically ahead when acceleration resumes.”
For investors, a controlled voluntary slowdown therefore carries a very different implication from a prolonged interruption that could be interpreted as the loss of technological leadership.
A Trillion-Dollar Investment Cycle
Samer Choucair said the scale of capital behind artificial intelligence makes any change in development cadence relevant far beyond OpenAI itself.
Goldman Sachs Research estimates that global AI-related investment could reach approximately $1 trillion in 2026, including about $581 billion in the United States. Consensus estimates cited by Goldman also place capital expenditure by major hyperscalers at close to $800 billion this year.
Choucair said the AI boom should no longer be viewed as primarily a software investment cycle.
It has become an asset-heavy infrastructure cycle encompassing advanced semiconductors, data centers, networking equipment, liquid cooling, electricity generation, grid infrastructure, and increasingly large-scale financing requirements.
That distinction matters.
A slower pace of frontier-model training could alter the timing of demand for the most compute-intensive training clusters, but it would not necessarily eliminate demand for inference, enterprise applications, AI agents, cloud infrastructure, cybersecurity, or the deployment of existing models across industries.
“Investors should separate the economics of training the next frontier model from the economics of deploying the models that already exist,” Choucair said. “Those are becoming two different capital-allocation stories.”
The Gulf Enters the Infrastructure Phase
Samer Choucair said the debate in the United States has direct implications for Gulf sovereign investors, which are moving from being primarily consumers and financiers of technology toward becoming owners of the infrastructure on which AI operates.
Saudi Minister of Communications and Information Technology Abdullah Al-Swaha said on September 12 that the Kingdom is working with the private sector to develop more than 14 gigawatts of computing capacity, with Saudi Arabia seeking to combine compute, capital, and customer access to attract AI companies.
At the same time, HUMAIN, the PIF-backed AI company, is expanding across the AI infrastructure stack. In August, HUMAIN and DataVolt announced that they were jointly developing 100MW of the first 360MW phase of an AI-ready data center project at Oxagon in NEOM.
In Abu Dhabi, G42 has also held exploratory discussions with potential investors over raising billions of dollars in external capital, although no final decision has been announced regarding the structure or size of a transaction.
Choucair argued that the Gulf’s strategic advantage does not depend on rapidly reproducing OpenAI or Anthropic.
Instead, its strongest position may lie in controlling the infrastructure, energy, financing capacity, and sector-specific datasets required to operate AI at scale.
“The Gulf does not necessarily have to win the frontier-model race to capture a meaningful share of the economics,” Choucair said. “Owning the compute, power, data infrastructure, and long-duration capital behind the race can itself become a strategic position.”
Risks and Opportunities
Choucair said one of the most significant risks is a classic coordination problem.
If one frontier laboratory slows while competitors continue scaling aggressively, the cautious company could risk losing market share, talent, technological leadership, or investor confidence. That makes coordinated safety standards economically difficult even when companies agree that greater caution is desirable.
Geopolitical competition represents another constraint. Chinese developers and increasingly capable open-weight models can make unilateral restraint more difficult for U.S. laboratories, particularly when artificial intelligence is increasingly treated as a strategic national capability.
Private-market valuations also face a potential repricing.
Companies valued on the assumption of uninterrupted exponential improvements in model capabilities could face pressure if investors begin extending development timelines or applying higher discount rates to future revenue expectations.
The effects could also spread down the infrastructure chain. A sustained reduction in the pace of frontier training could influence expectations for semiconductor suppliers, data-center developers, power producers, and other businesses whose valuations assume continuously accelerating compute demand.
Yet Samer Choucair said a coordinated slowdown could also create investment advantages.
Additional time could allow regulators in Washington and Europe to develop clearer safety frameworks while giving large AI companies the opportunity to build more sophisticated monitoring, containment, and governance systems.
In that environment, compliance itself could become a competitive barrier to entry.
The companies capable of meeting expensive safety, cybersecurity, testing, and reporting requirements may gain an advantage over smaller competitors unable to finance the same infrastructure.
Reordering AI Portfolios
Choucair said capital is unlikely to leave artificial intelligence altogether.
Instead, he expects part of the investment cycle to migrate from an overwhelming focus on raw model training toward the infrastructure and operating systems required to commercialize AI safely and at scale.
That means greater investor attention could shift toward compute infrastructure, electricity generation, grid capacity, cybersecurity, model monitoring, governance systems, enterprise applications, and sector-specific AI.
Choucair said investors should incorporate several possible trajectories into asset-allocation models through 2027.
One would involve selective slowing of frontier development while commercial deployment continues to accelerate. Another would involve more aggressive regulation that compresses valuation multiples across portions of the AI ecosystem. A third would see competition resume at maximum intensity if rival laboratories refuse to coordinate.
For Choucair, the most important conclusion is that the secular AI investment thesis has not disappeared.
“The structural trend has not been cancelled,” Samer Choucair said. “What has been cancelled is the illusion of a straight line.”
He added that investors able to balance exposure across computing infrastructure, power, compliance, cybersecurity, and commercial applications may be better positioned to capture returns from the next stage of the AI investment cycle than those relying exclusively on continued acceleration in frontier-model capabilities.