The race to dominate artificial intelligence has long been framed as a contest of scale—whoever spends the most on compute, talent and data should win. But Meta’s reported delay of its “Avocado” model, alongside discussions of licensing Google’s Gemini 3 technology, signals a turning point.
According to members of the Senior Executive AI Think Tank, the frontier of AI is becoming harder to sustain even for the most well-funded organizations. A recent analysis of Big Tech’s AI spending highlights how companies are pouring tens of billions into infrastructure while facing diminishing returns in performance gains—proving that capital alone is no longer enough to secure leadership.
This moment raises urgent questions for executives: If even hyperscalers struggle to keep up, what does competitive advantage in AI actually look like? And where does that leave smaller companies entering the race?
Below, Think Tank members attempt to answer these questions while looking toward what’s next. Together, their perspectives outline a new playbook for AI competition—one that begins with a surprising change at the very top.
From Builders to Buyers: The Rise of AI as Infrastructure
Rodney Mason, Chief Marketing Officer at Minty, says Meta’s willingness to consider licensing a rival model marks a profound strategic shift.
“A major shift—from builders to buyers (even at the top),” Mason says. “The most interesting part of the Meta example is that Meta might license a competitor’s model.”
He adds, “That’s a huge strategic shift: AI models are becoming commoditized infrastructure layers. Even the largest players may choose to simply plug into the best infrastructure over building from scratch.”
Drawing a parallel to cloud computing, Mason notes, “This is similar to how companies stopped building their own data centers and moved to cloud providers like Amazon Web Services.”
To Mason, the implication is clear: “Owning the product experience and data matters more than owning the AI infrastructure.”
Execution Over Scale: Why Speed Is the Real Differentiator
Mo Ezderman, Director of AI at Mindgrub Technologies, argues that Meta’s challenge is less about resources and more about execution.
“Staying at the frontier of AI is less about company size and more about speed, adaptability and decision-making,” Ezderman says. “Meta’s delay shows that massive investment alone isn’t enough without the right operating model.”
He contrasts this with competitors: “Google’s competitiveness highlights execution as the true differentiator.”
For smaller players, the implications are stark but clarifying. “The barrier is extremely high—training frontier models requires enormous capital, and incumbents already have a strong technical lead,” he says. “Smaller and open-weight models may get moments in the spotlight, but long term they’ll struggle to compete.”
Still, there is a path forward: “They will likely win by specializing in narrow, high-value use cases.”
Reports from leading research firms show that while large models dominate benchmarks, smaller, fine-tuned systems often outperform in specific enterprise applications—especially when paired with domain expertise.
“This moment says frontier AI is no longer just a capital race; it is a coordination race across talent, compute, data, product integration and speed of iteration.”
The Coordination Challenge: Why Complexity Is the New Bottleneck
Divya Parekh, Founder of executive coaching brand DivyaParekh.com, frames the issue as a coordination problem rather than a funding gap.
“This moment says frontier AI is no longer just a capital race,” Parekh says. “It is a coordination race across talent, compute, data, product integration and speed of iteration.”
She points to Meta’s situation as evidence: “Even giants can spend heavily and still lose tempo.”
For smaller companies, her advice is direct and actionable: “Do not outspend the hyperscalers. Win by being narrower, faster and closer to a real customer problem.”
Parekh emphasizes that success will come from focus, not scale. “The next wave belongs to focused builders, not just the biggest balance sheets.”
The Limits of Spending: Why More Compute Isn’t Enough
Chandrakanth Lekkala, Principal Data Engineer at Narwal.ai, highlights a growing disconnect between investment and outcomes.
“Frontier AI is becoming brutally hard to sustain, even with $135 billion,” Lekkala says. “The gap between training compute and actual benchmark performance is widening faster than spending can close it.”
He adds, “This makes it impossible to buy top positions by any company, no matter how big.”
For smaller companies, the takeaway is pragmatic: “It is practically impossible to compete directly with frontier models.”
Instead, he recommends a different path: “Specialization, efficiency and developing on existing models instead of building new ones.”
This reflects a broader shift toward efficiency-focused innovation, where improvements in architecture and deployment can deliver outsized returns compared with brute-force scaling.
“When you can’t outspend the incumbents, you outmaneuver them through robust modeling, efficient computation and open-source innovation.”
Innovation Over Scale: The Advantage of Mathematical Rigor
Dhyey Mavani, AI and Computational Math Researcher at Amherst College, takes a more technical view, focusing on the limits of scaling without foundational breakthroughs.
“Compute scales, but breakthroughs don’t follow a linear spending curve,” Mavani says. “Throwing hardware at a problem fails if the mathematical foundations lack elegance.”
He warns that “massive capital expenditures carry unprecedented risk if the underlying architecture isn’t fundamentally sound.”
Yet this challenge creates opportunity: “This is a massive opening for nimble teams.”
Mavani explains how smaller players can compete: “When you can’t outspend the incumbents, you outmaneuver them through robust modeling, efficient computation and open-source innovation.”
His perspective reinforces a key insight: Innovation at the algorithmic level—not just infrastructure—remains a powerful lever for competitive advantage.
The End of Permanent Leadership: AI as a Moving Target
Aditya Vikram Kashyap, Vice President of Firmwide Innovation at Morgan Stanley, sees Meta’s situation as evidence of a deeper structural shift.
“Scale alone is no longer a sufficient moat,” Kashyap says. “The limiting factor is not capital, but the ability to compound learning faster than competitors.”
He points to a changing competitive dynamic: “Frontier leadership is becoming episodic, not permanent.”
This has implications for strategy at every level. “Even the largest players are being forced into a more fluid build-buy-partner equilibrium,” he says.
For smaller companies, this creates a surprising opportunity: “The advantage will accrue to those who anchor themselves in proprietary data, narrow domains and faster execution loops.”
Rather than chasing dominance, companies can succeed by exploiting fragmentation in the market.
Competing to Survive: Speed, Pressure and Consolidation
Will Conaway, President of Tuxedo Cat Consulting, frames the moment in stark terms.
“Meta’s decision to pause its ‘Avocado’ AI and consider licensing Google’s Gemini—even after pouring $135 billion into AI—exposes a sobering reality: Deep pockets don’t guarantee leadership in this field,” Conaway says.
He highlights the pressure facing smaller firms: “The climb is even steeper. Soaring costs, technical hurdles and a relentless pace make it harder than ever to break through.”
Speed is critical. “Getting to market quickly is everything,” he says. “Many promising startups will be snapped up before they can compete alone.”
With new rules and ethical demands raising the stakes, his conclusion is blunt: “In today’s AI race, only those who adapt, partner wisely and deliver real results will survive.”
“Enterprise leaders should stop watching the model race and start building delivery capability that is model-agnostic.”
Model-Agnostic Leadership: Why Process Beats Hype
Markus Kopko, AI-PM Transformation Architect at Alvission Education GmbH, emphasizes operational maturity over model obsession.
“Meta’s situation is a model-builder problem. For most enterprises, it is irrelevant,” Kopko says.
He warns against overreliance on any single system. “If a company spending $135 billion cannot guarantee frontier performance,” he says, “betting your AI strategy on any single model is a structural risk.”
Instead, leaders should refocus: “Enterprise leaders should stop watching the model race and start building delivery capability that is model-agnostic.”
For Kopko, execution becomes the differentiator. “The companies gaining ground are not picking winners in the model wars—they are building repeatable processes that turn any capable model into measurable business outcomes.”
How Leaders Can Compete in the New AI Landscape
- Treat AI models as infrastructure, not differentiation. Focus on experience and data ownership.
- Prioritize speed and execution over scale. Agility beats investment alone.
- Win through focus and coordination. Align tightly around real customer problems.
- Avoid competing directly at the frontier. Build on existing models efficiently.
- Invest in smarter innovation, not just more compute. Breakthroughs require better thinking, not just bigger budgets.
- Exploit fragmentation in the AI market. Use niche expertise and proprietary data.
- Move fast and partner strategically. Speed to market and smart partnerships determine survival.
- Adopt model-agnostic strategies. Ensure flexibility regardless of which models lead.
The End of Scale as AI’s Ultimate Advantage
Meta’s AI delay signals a broader truth: The frontier of artificial intelligence is no longer controlled by scale alone. Execution, coordination and adaptability now define leadership.
For smaller companies, this shift is not a disadvantage—it is an opening. Success in AI will no longer be determined by who spends the most, but by who can focus strategically, move quickly and deliver measurable value.
