Qatar’s Ooredoo, Nvidia, and Nokia unveil multi-billion-dollar AI compute platform—and Southeast Asia is their target

Ooredoo Group plans to secure 1GW of AI compute capacity over the next three years via its Indonesia-focused Zankore neocloud platform, in partnership with Nvidia and Nokia (powered by GPU-as-a-Service). It has already lined up blue-chip demand for ~200MW for H1 2027, targeting ~$13B revenue and ~$9B cumulative EBITDA over 5-6 years, with Ooredoo’s proportionate EBITDA contribution around ~$600M; Ooredoo will invest $800M over five years (49% stake) to develop the platform. The initiative also highlights strong funding capacity (net debt/EBITDA ~0.6x vs. board guidance 1.5-2.5x), supporting a risk-on strategy to capture Indonesia/Southeast Asia’s growing AI data-center demand.

Analysis

This is less a near-term telco story than a long-duration option on Indonesia becoming a scarce AI utility market. The real economic question is not whether demand exists, but whether the first wave of capacity can be converted into durable contracted cash flow before competitors and power constraints normalize returns. If the tenant mix is genuinely creditworthy, the market should re-rate the asset base as infrastructure-like rather than telecom-like, which matters more for multiple expansion than for current earnings.

The clearest second-order winner is NVDA: every incremental sovereign/enterprise GPU cluster outside the U.S. reinforces vendor lock-in and expands the installed base for networking, orchestration, and software attach. But the bigger beneficiary may be INDO/Ooredoo itself if management proves it can finance growth while keeping leverage below board limits; that creates room for a multi-year capital cycle without forcing equity dilution. The losers are regional colo operators and hyperscalers facing tighter power availability in Southeast Asia, because scarcity shifts pricing power toward whoever controls land, power, and permitting rather than pure cloud brand.

The main risk is execution lag, not demand. These projects tend to trade well on announcement and then de-rate if power delivery, build timelines, or customer onboarding slip by 6-18 months. The contrarian view is that the market may be overpaying for the word 'AI' when the real bottleneck is still basic infrastructure; if utilization on the first 28 MW / 200 MW tranche disappoints, the equity story becomes a capex-heavy telco with lower ROIC, not a scaled AI platform.

Key Decisions for Investors

• Long INDO on any post-announcement pullback, with a 6-12 month horizon: thesis is multiple expansion from telco to infra/colo, but only if first-wave utilization and contracted revenue convert; exit if capex rises faster than EBITDA contribution or leverage moves toward management's upper guide.
• Buy NVDA only as a modest expression of the infrastructure buildout theme, preferably on weakness over the next 1-3 weeks; treat this as validation of global AI demand rather than a standalone earnings catalyst, and fade if GPU spending broadens without improving supply discipline.
• Watchlist alert: pair regional data-center/telecom beneficiaries against broad Asia tech if Indonesia AI capacity announcements accelerate but execution metrics lag; the best short is whichever local proxy first starts talking about power scarcity without disclosed contracted load.
• Set a 1-2 quarter catalyst check on INDO: if the initial capacity tranche does not show visible tenant ramp and financing terms stay equity-heavy, the trade should be reduced because the market will reprice this back toward a conventional telecom multiple.