• Post author:
  • Post category:AI World
  • Post last modified:February 15, 2026
  • Reading time:4 mins read

MENA VC surge, AI IPO thaw, and GLM-5: signs of a new build cycle

What Changed and Why It Matters

Capital is quietly rotating toward the next phase of the AI buildout. In early 2025, AI dominated venture flows and public-market appetite began to broaden from infrastructure to software.

EY’s latest IPO readout captures the shift:

“While the AI narrative dominated VC funding in the first half of 2025, highlighted by the US$40b Q1 investment in OpenAI, software IPOs beyond the CoreWeave …”

At the same time, the Gulf’s venture ecosystem is accelerating. A new regional roundup highlights that 2025 is a rebound year and asks the right question:

“Last year was a huge year for anything AI related. How has that thematic played out in the region as well?”

On the technology front, China’s Zhipu AI just pushed the large-model frontier while driving down inference cost:

“Zhipu AI launched GLM-5, an Opus-class model scaling from 355B to 744B parameters with DeepSeek Sparse Attention integration for cost-efficient long-context …”

Here’s the part most people miss: policy and power are now as central as model quality. Even local political debates are being reframed by data center plans and trade posture:

“The new AI data centres and U.S. trade talks drew attention away from issues that had been critical in early 2024.”

Zoom out and a pattern emerges: liquidity is returning, regions with dry powder are leaning in, the IPO window is cracking open, and the stack is shifting from “build the compute” to “ship products that monetize it.”

The Actual Move

  • VC flows: GCC/MENA investors are stepping up deployment across emerging markets, with AI still the main thread in 2025 funding dialogues.
  • Public markets: The IPO window is thawing for AI-adjacent software, not just core infra. EY flags broadening investor appetite beyond the usual GPU/compute names.
  • Model capability: Zhipu AI’s GLM-5 blends massive scale (355B–744B) with sparse attention for cheaper long-context inference—pointing to a global cost-down race.
  • Policy and infrastructure: Political focus is coalescing around AI data centers, power availability, and trade ties—now first-order variables for where capacity lands.
  • Macro sentiment: Weekly macro notes cite “AI capex” as a price-setter, while crypto discourse frames the liquidity backdrop with lines like:

“Bitcoin is not in a cycle. Fiat is. Bitcoin is the exit valve. If you zoom out …”

  • Industry enablers: Chip and wireless incumbents such as Qualcomm remain key in on-device and edge AI distribution, connecting the data center to consumer endpoints.

The Why Behind the Move

• Model

Sparse attention plus large parameter counts is becoming the default path to long-context performance without runaway inference costs. GLM-5 underscores that cost/perf is now a feature.

• Traction

As GPUs normalize, buyers want applications that turn tokens into revenue. That’s why software IPOs beyond core infra are reappearing: real workflows, clearer unit economics.

• Valuation / Funding

Capital is rotating to regions with sovereign dry powder and clear policy tailwinds (GCC/MENA) and to public markets seeking AI revenue exposure beyond infra concentration risk.

• Distribution

Edge and wireless channels matter. Incumbents in connectivity can unlock distribution at scale for AI agents and on-device inference—where latency, privacy, and cost win deals.

• Partnerships & Ecosystem Fit

Sovereigns, utilities, and hyperscalers are becoming one conversation. Winning requires alignment with grid timelines, land, cooling, and export/trade constraints.

• Timing

A reopening IPO window plus fresh VC appetite creates a brief lane to scale go-to-market. The cost-down wave in long-context models makes new products viable now, not later.

• Competitive Dynamics

The US still leads in frontier model adoption, but China’s cost-efficient scaling and MENA’s capital mobilization intensify competition on price, capacity, and speed to market.

• Strategic Risks

  • Power constraints can outpace model progress.

  • Public-market scrutiny punishes growth without margins.

  • Policy shifts (export controls, data residency) can stall deployments.

  • Liquidity whiplash—macro or crypto-driven—can snap risk appetite fast.

What Builders Should Notice

  • Cost beats novelty. Long-context only matters if it’s affordable in production.
  • Distribution is the moat. Edge, carrier, and OEM channels compound faster than marketing spend.
  • Choose capital by constraint. If power or policy blocks you, go where sovereigns clear the path.
  • IPO discipline is back. Clean unit economics and regulated-footprint stories travel better.
  • Design with the grid in mind. Power, land, and latency are now product features.

Buildloop reflection

The next AI winners won’t just train bigger models—they’ll compress cost, align with power, and ship where capital moves fastest.

Sources

Tags: , , , , , , ,