The AI Gold Rush Is Real And So Is The Irrational Exuberance

Forbes

3 days ago

Salman Khan, President & CEO of The Gideon Group.
The world is high on AI. From Sand Hill Road to the GCC sovereign corridors, everyone wants a piece of the generative intelligence revolution. But let's cut through the noise.
We are in a bubble, but not one born entirely of delusion. There's a structured wildness to this capital cycle that looks irrational only if you ignore the underlying megatrend. If the dot-com era was about digitizing information, this era is about digitizing cognition. That changes everything, including the nature of the bubble itself.
Internet 1999 Versus AI 2025
The similarities between today's AI boom and the late '90s dot-com bubble are striking. Both saw intense hype, outsized capital allocation and a herd mentality. But the scale has changed, and so has the substrate. In the 1990s, 'every business needs a website' became the mantra. In 2025, it's 'every company needs an AI strategy.' Valuations have soared, companies are scaling without solid monetization models and the narrative is being used as a vehicle for inflated funding rounds.
Just like the 1990s, we're seeing capital-chasing narratives, startups scaling without revenue clarity and talent hoarding before product-market fit. Only now, the numbers seem to be 10 times bigger, and the burn is faster.
The Anatomy Of Today's AI Bubble
AI companies are commanding multi-billion-dollar valuations without even a roadmap to sustainable revenue. This is a big red flag in my experience. The disconnect between genuine value and bloated valuations is real in 2025.
Most AI startups are simply API wrappers on top of foundational models like OpenAI or Meta. They lack their own large language models, defensible data flywheels or unique training architecture. Yet they're raising capital as if they're chipmakers.
The real gold rush is in GPUs. Amazon has earmarked over $100 billion for AI infrastructure. It seems to me that there's more heat in the compute market than in the models themselves.
We're watching the mass appointment of chief AI officers with vague mandates and no technical remit. It's reminiscent of 1999's 'e-something' craze. Boardrooms are pushing AI mandates into product roadmaps without a lack of understanding about cost, latency or inference margins.
Prompt engineering is the new day trading. Social media is saturated with self-appointed AI experts. Web3 influencers have rebranded overnight. YouTube channels and Twitter threads pump speculative large language models (LLMs) agents with zero use-case depth.
The Smart Capital Playbook
Despite all the hype around AI, the technology has played a role in reshaping logistics, pharma and cybersecurity. The technology itself is not the issue, but rather the timeline assumptions and our mentality around it.
Compute costs are still prohibitive for broad-scale commercial deployment. Monetization models are duct-taped, especially in consumer AI. Regulatory frameworks haven't caught up, and that introduces systemic drag. Patience and due diligence are required to play the long game of sustainable growth and development within the industry.
For leaders looking to navigate this AI gold rush, here's my strategic take:
• Integrate AI where it monetizes. Don't just implement where AI demos well. Consider stack exposure across LLM orchestration, inference optimization and agent memory systems. If it doesn't reduce cost, increase throughput or give you pricing power, it's noise. Don't pursue vanity projects.
• Forget building another LLM; consider owning the rails underneath. Infrastructure, orchestration, billing logic, compliance layers—that's where I believe the margins will sit. Consider capex-layer plays like chips, compute leasing and AI-ready data center infrastructure. The model race may not be the best pursuit unless you're sitting on sovereign-scale capital.
• Your data layer is your moat. Treat it like capital. You don't need better prompts. You need structured, compliant, high-fidelity data pipelines. AI eats that for breakfast.
• Optionality is the hedge. Vendor-lock can lead to stagnation. Stay multimodel, multicloud and modular. Anyone getting married to a single provider today may find themselves obsolete.
• Slow is smooth, and smooth is fast. You don't need to be first. You need to be resilient. Focus on building things that are inevitable five years from now, not impressive for five minutes. This wave will wipe out the loud and reward the precise. So play the long game. Own the rails. Stack the options. Monetize what others are still trying to understand.
Final Thoughts
Yes, we're in a bubble. But bubbles built railroads, fiber optics and mobile ecosystems. I believe that AI will be no different. What matters most is where you're positioned when this bubble bursts. Because this time, I believe some of these valuations will be vindicated. The rest? Burned for GPU fuel.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?