When Two Dashboards Tell Completely Different Stories
Why I remain extraordinarily bullish on artificial intelligence, but increasingly convinced markets are pricing the wrong economics.
For the past several years, investors have become accustomed to watching the same dashboard.
Bond yields.
Credit spreads.
Inflation.
Central bank policy.
Valuation multiples.
Liquidity.
Leverage.
Collectively, these indicators attempt to answer one question.
How much financial risk is building inside the system?
Today, that dashboard is becoming increasingly difficult to ignore.
• Put/call ratios
• Margin debt
• Bond yields
• CDS spreads
• Credit spreads
• Yen carry trade pressure
• US dollar liquidity (DXY)
• Inflation expectations
• FOMC guidance and rate expectations
• Dow Theory / Dow Transport divergence
• Shiller CAPE and other normalised valuation ratios
• Extreme market concentration
• 41% of the S&P 500 driven by the AI trade
• AI revenue customer concentration
• Circular financing within the AI ecosystem
• Trillions of dollars of off-balance-sheet obligations, including leases, power contracts and RPO commitments
• Peak speculative behaviour in private and public markets
• Rapidly falling token prices
• Markets continuing to shrug off deteriorating fundamentals
None of these indicators exist in isolation.
Together, they describe a market priced for extraordinary expectations.
Yet, at exactly the same time, another community is watching an entirely different dashboard.
The Open-Weight Dashboard
Researchers building open models are not obsessing over valuation multiples.
They’re focused on one question.
Can intelligence become cheaper?
Their optimisation loop is remarkably simple.
• Better models
• Open releases
• Community experimentation
• Distillation
• Improved architectures
• Smaller frontier-capable models
• Lower inference costs
• Lower token prices
• Wider adoption
• Repeat
Every successful iteration reduces the amount of computation required to achieve the same capability.
Every open release gives the next generation a better starting point.
Every optimisation compresses more intelligence into less computation.
The objective isn’t a higher valuation.
It’s a lower cost of intelligence.
Two Economic Loops
This is the contradiction I believe markets have yet to fully reconcile.
Financial markets continue valuing artificial intelligence as though scarcity remains the dominant economic force.
The open-weight community behaves as though abundance is inevitable.
One loop rewards increasing valuations.
The other relentlessly reduces the cost of producing intelligence.
One assumes scarcity.
The other manufactures abundance.
Both cannot remain true indefinitely.
If intelligence continues becoming dramatically cheaper to produce, the assumptions underpinning today’s infrastructure spending, pricing power and valuation multiples will inevitably come under increasing scrutiny.
This is not an argument against artificial intelligence.
Quite the opposite.
It is an argument that AI may become even more transformative than markets currently expect.
But transformational technology and transformational investment returns are not always the same thing.
History has shown this repeatedly.
Railways transformed economies.
Many railway investors lost fortunes.
The internet transformed commerce.
Many early internet leaders disappeared.
Technology can change the world while the economics evolve in unexpected ways.
Artificial intelligence may prove no different.
The Market May Be Watching The Wrong KPI
Perhaps the most revealing observation is that the open-weight community effectively has a single performance metric.
Can we make intelligence cheaper?
Everything else serves that objective.
• Better routing
• Better memory
• Distillation
• Continuous learning
• Improved caching
• More efficient architectures
• Smaller specialist models
• Open collaboration
• Inference optimisation
Different techniques.
One outcome.
Lower computational cost.
Financial markets, meanwhile, remain focused on a different set of metrics.
• Revenue growth
• Capital expenditure
• Infrastructure deployment
• Market share
• Pricing power
• Valuation multiples
Those metrics mattered when intelligence itself was scarce.
They matter differently when intelligence becomes abundant.
A Different Kind of Bull Case
This is where I think the debate is often misunderstood.
I remain extraordinarily bullish on artificial intelligence.
Perhaps more bullish than ever.
Every month the technology becomes more capable.
More efficient.
More accessible.
More useful.
The question isn’t whether AI succeeds.
The question is where the value ultimately accumulates.
As the cost of intelligence continues to collapse, competitive advantage shifts.
From producing intelligence…
…to organising it.
From proprietary capability…
…to trusted resolution.
From closed systems…
…to reusable structure.
From computation…
…to coherence.
That is a very different economic story.
Resolution
Financial markets and the open-weight community are watching two completely different dashboards.
One is measuring the growing fragility of financial markets.
The other is relentlessly measuring the falling cost of intelligence.
Eventually, those two dashboards will have to reconcile.
The technology can continue accelerating while the economics that markets have priced into today’s AI leaders evolve in a very different direction.
The most important question is no longer whether artificial intelligence will change the world.
It almost certainly will.
The more important question is whether financial markets have correctly identified where the value created by that transformation will ultimately settle.
Darren Swayne, 6th August 2026