In May 2026, Applied Digital’s stock surged by approximately 38% in a single month. During intraday trading on May 28, the stock hit a 52-week high of $50.72. However, by June, the market’s rhythm shifted noticeably—by the close on June 5, the stock had fallen to $39.62, down 10.26% for the day, with after-hours trading pushing it further down to $38.74.
This volatility isn’t unique to Applied Digital; it reflects a broader shift in the pricing logic across the entire AI data center infrastructure sector. Over the past 18 months, the market’s narrative premium for AI compute assets has largely hinged on a single factor: the "supply-demand gap." The four major hyperscale cloud providers plan to spend about $725 billion in capital expenditures in 2026, a 77% increase from $410 billion in 2025, with roughly 75% allocated to AI infrastructure. But in Q2 2026, this narrative is being challenged by at least three new variables: Blackstone’s BXDC entering the public market as a REIT, persistent upward pressure on energy costs, and a shift in rental pricing models from "capacity-driven" to "profit-driven."
Applied Digital’s June Pullback: After Surging Ahead, Fundamentals Come Under Scrutiny
Before diving into a structural analysis, let’s review the context behind Applied Digital’s pullback.
As of June 5, 2026, Applied Digital had delivered a year-to-date return of 61.58%, a one-year return of 210.26%, and a three-year return of 323.74%. Over five years, the return stood at 742.98%, compared to just 74.56% for the S&P 500. In other words, the stock’s outsized gains were concentrated over the past year and a half, rather than being steadily accumulated.
Referencing the company’s Q2 FY2026 financials (ending November 30, 2025), revenue reached $126.6 million, up 250% year-over-year. Data center colocation revenue was $41.6 million, up 15% YoY, with Jamestown (106 MW) and Ellendale (180 MW) operating at full capacity. High-performance computing colocation contributed $85 million, with about $73 million from CoreWeave’s leasehold fit-out services. Lease revenue recognized was roughly $12 million, with about $8 million in cash receipts.
However, revenue growth did not translate into improved profitability. Costs grew much faster than revenue: operating expenses rose 344% YoY to $100.6 million, with $69.5 million related to leasehold fit-out. SG&A expenses jumped 119%, mainly due to $23.8 million in accelerated stock-based compensation. Net loss was $31.2 million, or $0.11 per share—below consensus expectations. Operating cash flow was negative $97.9 million.
So why did the market continue to drive the stock higher after the earnings release (January 7)? The answer lies in the forward-looking information embedded in the narrative. The company disclosed contracts with hyperscale clients totaling 600 MW, expected to generate about $16 billion in lease revenue over the next 15 years. This figure is roughly 500 times the company’s trailing twelve-month revenue, exemplifying a "price driven by expectations." However, the drop from the May high of $50.72 to $39.62 on June 5—a 22% decline—was not event-driven. Instead, the market was recalibrating the "timeline for expectation realization"—specifically, how quickly that $16 billion in future lease revenue should be discounted into today’s share price.
Variable 1: Blackstone BXDC’s Entry—Valuation Benchmark Shifts from PE to REIT Logic
One of the deeper reasons behind Applied Digital’s pullback is the market’s shift in valuing AI data center assets—from the "high-growth tech stock" PE logic to the "stable cash flow asset" REIT logic. This transition wasn’t organic; it was triggered by Blackstone Digital Infrastructure Trust (BXDC) going public in May 2026.
BXDC raised about $2 billion, targeting the acquisition of completed, leased, and stabilized data centers in primary markets. Their tenants are investment-grade hyperscale clients under long-term triple-net leases. Target asset valuations range from $250 million to $1.5 billion, with long-term capitalization rates of roughly 5.75% to 7.0%.
Underlying market data helps explain the rationale. By the end of 2025, primary market vacancy rates had dropped to a historic low of 1.4%. Net absorption in 2025 reached around 2,498 MW, with Northern Virginia contributing 1,102 MW and Dallas 470.8 MW. In other words, BXDC isn’t building new assets—it’s acquiring infrastructure already validated by the market and generating stable cash flows.
Wall Street’s response to BXDC’s listing was mixed. RBC Capital rated it "Outperform" with a $24 target, highlighting its "differentiated entry point" and Blackstone’s deep relationships with hyperscale clients. JPMorgan gave a "Neutral" rating with a $23 target, noting that the REIT structure is better suited for long-term holding, but current valuations offer no clear discount relative to net-lease peers. BMO also rated it "Market Perform," emphasizing that success depends on capital allocation efficiency and funding cost management, while expressing caution about the "external management structure" and "blind pool status."
BXDC’s entry impacts Applied Digital by providing a "benchmark comparable asset" for the market. Before BXDC, AI data center assets lacked a transparent, comparable public market pricing benchmark. BXDC is priced at a capitalization rate of about 5.75% to 7.0%, while Applied Digital is still valued on an EV/Revenue basis (currently about 38.9x). When the same underlying asset—AI data centers leased to hyperscale clients—faces two entirely different valuation methodologies in the same public market, cross-asset comparison and arbitrage inevitably follow. This cross-category comparison creates structural valuation pressure for "tech narrative" stocks like Applied Digital.
Variable 2: The Combined Effect of Energy Costs and Leasing Models
The pricing logic for AI data center infrastructure faces a second structural variable: energy costs.
Data center colocation isn’t simply a "landlord" business. Operators must bear electricity procurement costs while charging hyperscale clients for capacity. In Q2 FY2026, Applied Digital’s data center colocation revenue was $41.6 million, but operating profit for this segment was about $16 million, based on $130.8 million in operating assets. This profit margin is sustainable under full-capacity assumptions, but only reflects a static cost structure.
Marginal changes in electricity prices directly compress EBITDA margins. Using the 2025 North American data center PPA (long-term power purchase agreement) average of $55–$65/MWh as a benchmark, a 10% increase in electricity prices could cut EBITDA by 200–400 basis points under fixed-rent agreements. More importantly, AI data centers typically use high power density designs (30–100 kW per rack, compared to 5–10 kW for traditional centers), meaning power intensity per square foot is 3–10 times higher, amplifying the impact of electricity price fluctuations.
In contrast, BXDC’s lease structure provides a different buffer for energy costs. Triple-net leases require tenants to cover all operating expenses, including electricity, with landlords only collecting rent. With fixed annual rent escalators (typically 2%–3%), BXDC’s operating profits are far less sensitive to power costs than the colocation model. The difference in how these two models handle energy costs reflects fundamentally different risk-return preferences—colocation models participate in operational upside when energy costs fall, but bear concentrated pressure when costs rise.
The Broader Context: The Breadth and Depth of AI Infrastructure as an Investment Theme
Applied Digital and BXDC represent just two slices of the AI data center infrastructure investment landscape. Broadening the view, the theme spans several asset classes.
At the semiconductor level, Marvell Technology posted $2.418 billion in revenue for Q1 FY2027, leveraging partnerships with NVIDIA on NVLink Fusion and custom XPU solutions to transition from a cyclical chip supplier to a strategic AI data center player. In power infrastructure, Quanta Services is benefiting from grid modernization and hyperscale data center power access demand, with revenue projected to grow from about $35 billion in 2026 to $43.5 billion by 2028.
In the data center REIT space, Equinix operates around 280 data centers globally, serving over 10,500 enterprise clients—about 60% of which are Fortune 500 companies. Keppel DC REIT is shifting toward hyperscale and AI inference workloads, with hyperscale clients now making up 69.3% of its tenant base. In 2025, rent repricing rose 45%, and portfolio occupancy reached 95.8%. These figures highlight how data center REITs are benefiting from long-term structural trends—not just AI training cluster demand, but also AI inference, cloud migration, and digital transformation.
However, the industry faces a key structural issue: limited market share translates into highly concentrated bargaining power. The four major hyperscale clients (Amazon, Google, Meta, Microsoft) determine the sector’s pricing power. When over 70% of incremental demand comes from just four clients, any service provider—whether a colocation operator, data center landlord, or power infrastructure supplier—inevitably faces constraints from buyer concentration. This is a critical factor investors must incorporate into long-term valuation models for AI data center infrastructure assets.
Gate Launches Stock Trading: Understanding Thematic Investing Through "One-Stop Asset Allocation"
For investors focused on the AI data center infrastructure theme, a practical question arises: how can you build a cross-asset portfolio with minimal friction?
Traditionally, this involves multiple accounts and funding channels: a brokerage account for stocks like Applied Digital (APLD) or BXDC, a crypto account for digital assets, and if you want to add Marvell (MRVL), Equinix (EQIX), or other related companies, you’ll need to transfer funds and manage positions across several brokers. This fragmented account structure creates significant operational costs for investors seeking dynamic rebalancing within a theme.
On June 1, 2026, Gate officially launched real stock trading services, marking a major shift from "crypto exchange" to "multi-asset allocation platform." This service is enabled through a strategic partnership with regulated broker Alpaca, connecting directly to brokers with US Broker-Dealer licenses and clearing capabilities, and offering access to major exchanges including NYSE, Nasdaq, NYSE Arca, NYSE American, and BATS.
In terms of coverage, Gate Stocks supports over 10,000 stocks and ETFs—far more than most tokenized stock platforms, which typically offer only a few hundred assets. Beyond tech giants, it includes small-cap stocks and thematic ETFs across sectors. The service supports fractional shares as low as 0.01, with investments starting around $1, and settlements in USDT. Crucially, stocks purchased on Gate are independently custodied through the DTC system, with every position backed by real registered shares. During the holding period, users automatically receive full shareholder rights, including cash dividends, stock splits, and rights issues, with no funding rates or overnight holding fees.
This positioning gives Gate’s stock trading service potential value within the AI data center infrastructure theme. With a single account and USDT, users can participate in both digital asset and traditional stock markets. Looking ahead, as intraday trading expands to 24/7 and instant broker-to-broker transfers become available, the service could further reduce the total cost of cross-market asset allocation.
Risk Analysis: Ongoing Divergence in Asset Attributes and Pricing
To summarize the risk logic outlined in this article:
First, asset class identification risk. Applied Digital’s colocation model and BXDC’s REIT leasing model both involve AI data centers as underlying assets, but their revenue and cost structures are entirely different. As the market recognizes these differences, their valuations are likely to diverge. Investors need to clarify which risk-return profile they are choosing before picking individual stocks.
Second, energy cost risk. The profitability of the colocation model is highly sensitive to electricity costs, and North American power prices for 2025–2026 remain uncertain. If data center PPA averages break above $65/MWh, colocation operators’ EBITDA contraction could exceed market expectations.
Third, bargaining power concentration risk. Changes in the procurement decisions of the four hyperscale clients can have nonlinear effects on sector demand. Even if long-term AI compute demand remains strong, short-term order fluctuations can significantly impact the quarterly results of related companies.
Fourth, valuation regime shift risk. As discussed, the introduction of REIT logic brings a new valuation framework to AI data center infrastructure. This shift may continue to pressure companies still valued as tech stocks, especially if overall market risk appetite contracts.
Conclusion
Applied Digital’s 38% surge in May followed by a June pullback isn’t an isolated stock event—it’s a microcosm of the shifting pricing logic across the AI data center infrastructure sector. Blackstone’s entry into the public market with BXDC’s REIT structure offers investors a new benchmark focused on cash flow stability and asset scarcity. Meanwhile, the marginal impact of energy costs on colocation profitability, the constraints of buyer concentration on industry pricing power, and the divergence in valuation logic between business models collectively form the core investment framework for this theme.
Over a longer horizon, supply-demand dynamics for AI compute infrastructure will likely remain tight. But in public markets, asset pricing is never determined solely by fundamentals of supply and demand—it also depends on "how you value it" and "what you compare it to." At this stage, data center infrastructure is undergoing a cognitive shift from "narrative assets" to "cash flow assets." Only assets with the following characteristics are likely to remain competitive under the new valuation framework: stable lease structures, controllable energy cost exposure, and a diversified client base.
For investors seeking systematic exposure to this theme, Gate’s launch of real stock trading lowers the operational barriers and costs of cross-market allocation. With a single account, you can allocate across digital assets and US equities—not just to "diversify risk" or "dilute concentration," but to enable dynamic comparison and selection across business models. In thematic investing, the most valuable skill isn’t just "picking the right track," but always being able to choose the best segment within the track.
