Compute as Capital: Redefining AI's New Currency | WEKA, Akamai, TensorWave & More | RAISE 2026

Market Scale and Financial Dynamics

• A McKinsey report projects $7 trillion in global spending on AI infrastructure by 2030.
• Current AI infrastructure expenditure is approaching $300 billion in capital expenditures (CapEx), with potential derivatives and forward markets expanding this to a $3 trillion trading scale.
• Participants agree "compute" has replaced electricity and the internet as the primary catalyst of the third industrial revolution.
• H-100 GPUs are an anomaly in asset classes, appreciating ~20% year-over-year rather than depreciating as standard IT equipment does.
• Compute is redefining financial asset classes, with the CME launching options for compute volatility and "Compute Change" acting as a marketplace for token and compute trading.
• Val Bercovici (Weka) identifies "tokens" as the most liquid currency in the industry, serving a role similar to the U.S. dollar in traditional economies.

Primary Scarcities and Bottlenecks

• Power/Energy: Identified by all panelists as the single most critical constraint; access to gigawatts of grid power is harder to secure than hardware.
• Talent: Specific expertise in scaling data centers is scarce, taking decades to develop; White Fiber acquired a team with 20 years of hyperscaler data center experience to bridge this gap.
• Supply Chain: Physical resources including space, storage, and access to capital are now bottlenecks alongside GPU availability.
• Community Resistance: Data centers face increasing local opposition (NIMBYism) and regulatory moratoriums due to concerns over land use, noise, and environmental impact.
• Efficiency Bottlenecks: The industry is shifting from "GPU maxing" to "memory-bound" inference efficiency; optimizing the memory wall can generate $10 in value for every $1 of compute capital.

Business Models and Valuation

• Token Economics: Token costs per unit are decreasing, but total market consumption is rising; long-term contracts (36 months) are flattening in price, signaling a move toward market equilibrium.
• Value Shift: Current value capture is inverted, residing in the infrastructure supply chain (chips, power, data centers) rather than the application or SaaS layer.
• Forward Outlook: Value is expected to migrate up the stack by 2030–2031 as manufacturing capacity increases and supply chains stabilize.
• Refurbished Market: The resale value of refurbished H-100s is rising, challenging traditional depreciation models for hardware.
• Financial Intermediation: Panelists argue that Wall Street and banking involvement (via CME and central clearing) is essential to lower the cost of capital and manage counterparty risk.

Technological Innovations and Strategy

• Virtual Superclusters: White Fiber announced a new technology to link two disparate data center locations (within 83km) via dark fiber to create a virtual supercluster, bypassing single-site bandwidth constraints.
• Retrofit Strategy: White Fiber is converting abandoned industrial sites (e.g., mattress factories) into Tier 3 data centers to reduce construction time from two years to six months.
• Domain Specificity: The industry is moving away from generic "token maxing" toward fine-grained, domain-specific models driven by economic necessity and safety requirements.
• Global Distribution: Compute markets are expanding beyond the U.S. to Europe, Southeast Asia, and South America to bypass regional restrictions and leverage local energy.

Regulatory and Geopolitical Risks

• Government Intervention: U.S. and Chinese governments are implementing controls on powerful AI models, including export restrictions and caps on unregulated model distribution.
• Open Source Ceiling: A new ceiling is forming for open-source models, potentially limited by three-month release cycles or export controls.
• Social License: The lack of clear communication from AI leaders regarding human workforce impact has fueled public fear, making community relations a critical operational hurdle.

Panelist Forecasts and Future Scenarios

• Bubble Debate: Panelists reject the "AI bubble" narrative, characterizing current demand as a structural shift where supply will struggle to keep pace for years.
• Long-Term Asset: Sam Tabar (White Fiber) predicts power will be the most sought-after commodity in the medium-to-long term.
• Efficiency Focus: Carmen Lee (Silicon Data) notes that market discipline and short-selling are healthy mechanisms preventing extreme overvaluation.
• Success Metrics: Success for organizations like Weka and Akamai is tied to the ability of large enterprises to prove repeatable, scalable ROI across diverse verticals.
• Potential Disruption: The industry risks significant setbacks if data center moratoriums snowball or if geopolitical restrictions severely limit model availability.