Here's Why You Should Hold on to GDS Holdings Stock for Now

GDS Holdings has fallen 29.8% over six months, but consensus projects 2026 revenue of $1.9 billion, up 17.4% year over year, and EPS of $2.23, up 1,073.7%. AI-driven demand supported record 470 MW of bookings in H1 2026 and 600 MW of additional customer reservations, with management targeting 1 GW of full-year sales. Revenue visibility is supported by 89.2% pre-commitment on capacity under construction, but Q2 EBITDA margin fell 220 bps to 45.5% due to utility costs, while higher capex and sizable debt keep leverage risk elevated.

Analysis

GDS is transitioning from a demand-constrained to a power-and-capital-constrained business. The key valuation question is not whether contracted MW converts to revenue, but whether incremental revenue produces enough EBITDA and free cash flow to justify another funding cycle. A 220bp margin decline while utilization is still below full operating leverage suggests utility-cost pass-through and customer contract escalation clauses are more important than headline bookings; without them, each incremental MW can widen rather than relieve leverage concerns.

The next 1-3 month catalyst is evidence that the booking target converts into signed contracts with deposits, utility allocation, and construction financing—not additional reservations. Over 6-18 months, GDS could re-rate if backlog conversion lifts EBITDA faster than capex and net-debt/EBITDA declines; the adverse path is an equity or expensive offshore/onshore refinancing before cash generation inflects. China AI infrastructure demand also has a non-obvious concentration risk: hyperscalers can reserve capacity defensively, then stagger deployment if model monetization disappoints or power availability tightens.

Competitive read-through is mixed for Chinese data-center peers such as VNET and for global AI-infrastructure proxies. GDS's capacity commitments may indicate durable regional demand, but the more likely near-term beneficiary is the power-equipment and grid chain rather than the landlord: scarce power interconnection raises project cost and delays revenue recognition. FIGR and IT are not economically linked enough to GDS's operating outlook for a fundamental relative-value trade despite their inclusion in the source.

Consensus may be overemphasizing the large percentage EPS growth figure, which is flattered by a low base and does not resolve the equity value leakage from capex, interest expense, and potential dilution. The stock decline creates optionality only if management demonstrates stable EBITDA margin and funding discipline. Falsify a cautious stance if two consecutive quarters show margin stabilization above 45%, utilization rising toward the mid-80s, and capex fully covered by committed financing without a material increase in leverage.

Key Decisions for Investors

• Maintain GDS as watchlist/neutral rather than initiate on the share-price drawdown. Upgrade only after the next two earnings reports confirm backlog-to-billable conversion and EBITDA-margin stabilization; avoid treating reservations as equivalent to revenue.
• If expressing a bearish 3-6 month view, use a defined-risk GDS put spread rather than an outright short, sized small: the primary downside catalyst is a capex or refinancing revision, while strong AI-capacity headlines can drive sharp short-covering rallies. Exit if management funds the expansion without incremental leverage or dilution.
• For a long thesis, wait for a financing and power-availability disclosure, then enter GDS in tranches with a 6-12 month horizon. Target a re-rating only if utilization and margin both improve; cut exposure on another material margin decline or a reduction in contracted-capacity conversion expectations.
• Monitor VNET and Chinese hyperscaler capex disclosures as confirmation indicators. A broad acceleration in deployed—not merely reserved—capacity supports GDS; delayed cloud capex, regional power restrictions, or rising utility tariffs would favor avoiding the entire China data-center landlord cohort.