Morgan Stanley TMT Conference 2026: AI infrastructure and ‘Sovereign Data’ take center stage

The Morgan Stanley Technology, Media & Telecom (TMT) Conference, held from March 3 to 5, 2026, served as a critical barometer for the "second wave" of the AI revolution. While 2025 was defined by the race to build large language models, this year’s flagship event shifted focus toward the physical and economic bottlenecks of AI: power hunger, sovereign data residency, and the urgent need for a "product-led" leadership transition across the enterprise software sector.

The conference, which hosted giants like Meta, Cisco, and CrowdStrike, underscored a growing market reality: compute demand is now structurally outpacing supply, and the "energy bottleneck" has moved from a technical hurdle to a primary political risk.

The Energy ‘Wall’ and ‘Bring Your Own Power’

One of the most sobering themes was the looming power crisis. Morgan Stanley analysts and industry CEOs warned that U.S. data center demand could reach 74 GW by 2028, but current grid projections show a massive 49 GW shortfall.

• Off-grid independence: A recurring theme was the rise of "Behind-the-Meter" power. Hyperscalers are increasingly looking to bypass traditional utilities by building their own microgrids, utilizing natural gas, fuel cells, and small modular reactors (SMRs).

• The politics play: Policy experts noted that "AI electricity inflation" is becoming a midterm election issue. Candidates are beginning to campaign against data center buildouts that threaten to drive up residential utility bills, leading to calls for "data center exit fees" to protect household consumers.

• Pricing premium: Research shared at the conference indicated that corporations are now willing to pay a $20/MWh premium for power that is both "green" and "rapidly accessible," effectively decommoditizing the energy market.

Meta’s ‘Durable Tent’ and Efficiency Drive

Meta CFO Susan Li provided a rare look into the company’s "scrappy" approach to maintaining its AI lead while navigating massive capital expenditure requirements.

• Infrastructure hacks: To bridge the gap until its next-generation data centers go online in 2027, Meta is deploying "durable tents", rapid-build, high-performance temporary enclosures, to increase capacity for its Llama-series models.

• The iREV Metric: Meta is doubling down on iREV (Internal Revenue per Engagement), using AI to compound small gains in ad performance. Li reported that algorithmic improvements led to a 3% conversion lift on Instagram last quarter alone.

• In-house silicon: Meta is aggressively expanding its custom silicon efforts, moving beyond inference to use its own chips for model training, a direct attempt to reduce its multi-billion-dollar dependency on external chipmakers.

Cybersecurity as the AI winner

Amidst turmoil in traditional enterprise software, which saw a $1 trillion market cap loss in a single week in early 2026, cybersecurity emerged as the clear beneficiary of the AI era.

• Platform consolidation: CrowdStrike reported record net new ARR of $330.7 million, fueled by its "Charlotte" AI tool, which saw utilization increase 6x year-over-year.

• The identity perimeter: Okta executives emphasized that in a world of "autonomous agents," identity management is no longer just about humans. Securely managing the credentials of AI agents has become the new front line of enterprise security.

• The "product CEO" trend: Boards are increasingly replacing sales-oriented CEOs with product-oriented leaders who understand the technical architecture required for "AI-native" transformation.

The ‘File System for AI’

As businesses struggle to move past the "pilot phase" of AI, companies like Box and IBM are positioning themselves as the foundational "memory" for these systems.

• Unstructured data: Box CEO Aaron Levie argued that "files are the natural unit of work for an AI agent," noting that the company’s Enterprise Advanced plan now accounts for 10% of revenue as firms rush to clean up their data for AI consumption.

• Hybrid Cloud & AI: IBM’s Rob Thomas highlighted that the company’s focus on hybrid cloud, an area Arvind Krishna pivoted to in 2020, is now perfectly aligned with the "sovereignty" trend, allowing firms to run models on-premise without exposing sensitive data to public clouds.

"The upcoming infrastructure cycle will create islands of wealth and literal power," said Stephen Byrd, Morgan Stanley’s Global Head of Thematic Research. "We are moving into an era where access to a transformer and a turbine is the new competitive moat."