Behind the AI Investment Boom, a Governance and Cost Reckoning

Silicon Valley’s investment in artificial intelligence (AI) has reached substantial levels, including trillion-dollar projections, major infrastructure financing deals, and a compute buildout racing to keep pace with demand that may or may not materialize. By several financial measures, 2026 is shaping up to be a significant year for AI investment.

But behind the record checks, a quieter reckoning is taking shape. Security teams are discovering that autonomous agents can become liabilities as easily as assets. Some CFOs are approving AI spending that can be difficult to connect directly to business outcomes. And a growing chorus of skeptics is asking whether the industry is building infrastructure or inflating a bubble.

The money is real. Whether the returns and the risk management can keep up is the question.

Heavy on investment, but are we blowing a bubble?

Global investment in AI has increased considerably. Recent predictions by Goldman Sachs forecast global AI investment to exceed $1 trillion in 2026, including $581 billion in the United States alone. Large investments have recently emerged from companies like Nvidia; they are assembling up to $500 billion in financing with Apollo, Blackstone, BlackRock’s GIP, Brookfield, Goldman Sachs, and KKR to fund AI infrastructure and to mobilize third-party capital for the buildout of AI infrastructure over time.

“NVIDIA has reached an important milestone. We began by building chips; today, we are helping create a new class of productive, investable infrastructure: AI factories,” said Jensen Huang, founder and CEO of NVIDIA in a company release. 

“In AI, compute is revenue. NVIDIA compute is uniquely suited for this role. It is broadly adopted, flexible across models and workloads, fungible and transferable across customers and operators, and continuously improved through CUDA software — extending its useful life and improving its economics over time. It is supported by a deep global ecosystem of developers, customers and offtakers. That is why we are bringing the world’s leading long-term capital providers together to independently underwrite AI infrastructure. These financing platforms will help customers access scarce compute at scale and build the DSX AI factories that will power every industry and country in the age of AI,” shares NewRocket Chief Growth Alliance Officer Jason Rosenfeld.

Despite heavy investment, questions emerge and concerns compound among people and businesses alike. Questions about a potential “AI bubble” persist, including whether current financing arrangements are distributing infrastructure risk or contributing to demand that may depend on future monetization.

Chief Growth Alliance Officer Jason Rosenfeld says the financing structure may keep much of the associated credit risk off Nvidia’s balance sheet while supporting continued demand for GPUs. 

Exploring concerns with AI

Financial risk isn’t the only thing keeping executives up at night. Among the various concerns people voice about AI, one of the bigger ones weighing on the minds of executives and shareholders is a matter of how and when AI is being deployed, what the cost is, and the various cyber risks associated with deploying AI agents.

IBM found AI-enabled breaches cost organizations $6 million on average, exposing gaps in vulnerability management, access controls, and AI governance. Organizations using AI extensively in security operations saved nearly $2 million per breach; however, only 18% apply AI agents to vulnerability management, even though half already use them for detection. 

This gap is one that concerns organizations, and many AI-driven organizations are implementing strategies to combat and close it.

“We are deploying both Agentic AI and Security Operations solutions for clients. Autonomous security solutions are critical for proactive detection. As more and more companies deploy agents into their workforce they will become another source of vulnerability. Between this and the speed at which AI hackers will exploit vulnerabilities, you have both internal and external forces at hand which could potentially cripple a company if not addressed quickly enough. The speed at which this must be done can only be accomplished by utilizing AI to address vulnerabilities,” shares Rosenfeld.

Additional governance issues continue to emerge, and one that recently made headlines centered around the new phenomenon of rogue AI agents. Recent cybersecurity evaluations have highlighted the possibility of AI agents accessing systems or networks beyond their intended testing environments. Other internal evaluations have reportedly identified unintended pathways into external systems, illustrating the governance challenges associated with increasingly autonomous tools.

“The moment you deploy an agent that makes decisions, calls systems, and produces outputs that people act on, you have hired someone. Except you skipped everything you would normally do before letting someone loose in your organization. No defined scope of authority. No owner is accountable for its behavior. No review cycle. No audit trail,” explains BotGauge CEO Pramin Pradeep.

A recent Dark Reading poll found that roughly half of cybersecurity professionals now rank agentic AI as the top attack vector for 2026, ahead of ransomware and deepfakes. 

The hidden costs with AI

That absence of accountability does not remain confined to security. As organizations and executive leadership teams are acutely aware, it shows up clearly on the balance sheet.

Other concerns commonly voiced center around the cost versus benefit of implementing expensive AI, and how it is even billed. Ninety-five percent of enterprise generative AI pilots produced no measurable P&L return, against $30 to $40 billion in enterprise investment. Additionally, a Deloitte CFO AI guide reports that more than half of respondents now allocate between 21% and 50% of their digital initiative budgets to AI, yet many CFOs can’t trace that spending to specific business outcomes.

The cost of AI implementation and running various AI programs is not a bill that comes once. In reality, it is driven by token costs. Output tokens typically run 3x–4x more expensive than input tokens, which means a bloated prompt costs you twice: once for the unnecessary input, and again for the lengthy response it tends to trigger. 

“Too many CFOs in 2026 received their bills and realized that token usage was way higher than they predicted. I don’t think that CFOs understood the difference in the models and how that impacted consumption, nor did most people within companies. Now more than ever it’s critical that end users are educated on when to use AI and when not to use AI, as well as which models to use for what purposes. I have seen too many organizations have incredible bills just because people were asked to consume AI but were not educated on how. The bill itself is only part of the surprise. What catches people off guard is the data cleanup, the integration work, the tooling needed to govern any of it, and the hours their own people spend reviewing and reworking output that wasn’t right the first time,” Rosenfeld says.

Organizations today may be managing subscriptions, usage-based billing, performance-based contracts, hybrid structures, freemium tiers, and agentic seat pricing, sometimes across multiple tools simultaneously. Microsoft Copilot appears to cost $30 per user per month, but that price assumes an existing Microsoft 365 subscription, making the true all-in cost substantially higher. Token bloat is what happens when an AI system processes far more information than it actually needs. It’s one of the most common, and least visible, sources of runaway AI costs.

Meanwhile, AI has pushed software prices up by 20–37% across vendor categories, although customers may not always receive a corresponding increase in functionality. Some researchers are calling this an “AI tax.” Costs continue to rise while revenue growth may not keep pace.

How can companies, who are struggling with tighter budgets largely due to AI scaling and implementation, save their marketing budgets from hypotheticals? And do these questions about hidden costs make companies more conservative?

Rosenfeld weighs in, “I wouldn’t necessarily say businesses are becoming more conservative, but I would instead say they are becoming smarter when it comes to AI implementation. They are asking for more visibility, more control and more accountability. No one can deny the power of using AI, but companies now want to see proof of what they got back for the money they spent. That is the same discipline they already apply to every other line item in the budget, and it is finally getting applied here, too.”

Who comes out on top of the AI race?

If spending without accountability is the problem, the winners will be defined by who solves it first. As the AI arms race continues at full throttle, we will see which moves leaders in the space make prove successful, and, perhaps more importantly, which do not.

“The winners will not be the ones who spent the most on AI. They will be the ones who understood what they were actually deploying. History suggests superior technology is merely the entry ticket, not the key to enduring leadership and the AI race is proving this out in real time. The companies pulling ahead share one characteristic, they treat AI as an operational discipline, not a procurement decision. They know which agents are running in their systems, what those agents are doing, and whether the outcomes they are producing match the outcomes they were designed for the market use cases. That sounds basic. Fewer than a third of enterprises currently meet that standard,” shares Pradeep.

For executives and organizations, success may depend partly on treating AI as a managed investment rather than a speculative bet at a casino.

“The ones treating AI as a bet they placed and are waiting to pay off will be disappointed,” Pradeep states.

Gartner predicts more than 40% of agentic AI projects will be cancelled by the end of 2027, a number that suggests the reckoning described throughout this piece is already underway. The capital flooding into AI infrastructure is not in question. 

“The organization that continues to invest in AI, but does so wisely, and not blindly. Too many organizations in 2026 were running without the right processes and tools in place to manage, while simply just telling their people to start using it. These organizations need to ensure they have controls in place to manage spend across their company, identify and align value from their spend, educate their workforce on when and how to properly use AI, and tie every use case back to a business outcome that somebody actually owns. The companies that come out ahead will be the ones that can tell you exactly what they got for their spend,” explains Rosenfeld.

What remains unresolved is whether the industry spending it will build the governance, cost discipline, and operational rigor to match, or whether the gap between ambition and management will simply compound until it is too late to close.

The information provided in this article is for general informational and educational purposes only. It is not intended as financial advice. Readers should not rely solely on the content of this article and are encouraged to seek professional advice tailored to their specific circumstances. We disclaim any liability for any loss or damage arising directly or indirectly from the use of, or reliance on, the information presented.      

Prices and availability are accurate as of the time of publication and are subject to change without notice. Please check the retailer’s website for the most up-to-date pricing information.

Entrepreneur Media was not involved in the creation of this content.