Monolithic Power Systems posts record $980m Q2 on AI data-centre surge

MPS reported Q2 2026 revenue up 47.6% year on year, driven by a 164% annual jump in Enterprise Data sales.

Electronic circuit boards densely wired with multicolored cables are mounted in a metal frame within a brightly lit cleanroom.

Monolithic Power Systems (MPS) posted record quarterly revenue of $980.6 million in the second quarter of 2026, beating the prior quarter by 21.9% and the year-ago period by 47.6%. The fabless power-semiconductor specialist said the result was driven by broad-based ordering across all six end markets, with Enterprise Data as the standout performer.

Enterprise Data revenue reached $380.6 million in the quarter, up 44.8% from Q1 and an striking 164.3% higher than Q2 2025. The segment now accounts for 38.8% of total revenue, compared with 32.7% just one quarter earlier. MPS attributed the acceleration to higher sales of power management solutions for AI servers and associated infrastructure, reflecting the sustained build-out of hyperscale and enterprise AI compute capacity globally.

Financial highlights

GAAP net income for the quarter came in at $257.3 million, or $5.22 per diluted share, against $135.0 million a year ago, representing 90.6% year-on-year growth. On a non-GAAP basis, net income was $320.1 million, or $6.50 per diluted share, up 58.3% year on year. GAAP operating margin expanded to 31.0%, up 6.2 percentage points year on year, while non-GAAP operating margin reached 37.5%.

Cash and short-term investments stood at $1.41 billion at the quarter's close. The company also announced that its board has authorised an additional $500 million for share repurchases, lifting the total current buyback programme to $1 billion.

For Q3 2026, MPS guided revenue in the range of $1.14 billion to $1.16 billion, which would represent a further step-up of roughly 17% at the midpoint. GAAP gross margin is guided at 55.2% to 55.8%.

"Our results demonstrate the strength of our diversified model and our continued success in transforming from a chip-only, semiconductor supplier to a full service solutions provider," said Michael Hsing, CEO and founder of MPS.

The company also disclosed that it received initial orders for high-speed DDR5 memory components and began sampling high-voltage AC-to-DC products targeted at 800V data-centre power architectures. MPS described these moves as part of its strategy to expand its serviceable addressable market beyond current AI and server core power positions.

Market context

MPS competes in the analogue and mixed-signal power management space alongside Texas Instruments, Renesas, Infineon and a cluster of smaller fabless vendors. The AI infrastructure build-out has become the primary demand driver across the sector, with power delivery at the rack and chip level emerging as a critical engineering constraint as GPU clusters scale toward multi-megawatt densities.

The 800V data-centre architecture work is particularly significant: as AI accelerator clusters draw more power, traditional 48V power delivery topologies are being pushed toward higher-voltage architectures to reduce copper losses and improve energy efficiency. MPS's early sampling activity in this area positions it ahead of the broader product cycle, though the company did not disclose customer names or volume commitments.

Automotive remained the second-largest segment at $157.1 million, up 8.2% year on year, with MPS reporting products shipped across more than 1,500 new sockets so far in 2026, primarily in ADAS and infotainment applications. This diversification is material: should AI infrastructure spending moderate, automotive and communications (up 78.3% year on year, driven by optical modules and switch demand) provide meaningful secondary growth vectors.

Inventory days fell from 157 at the end of Q1 to 140 at the end of Q2 on a current-quarter basis, a constructive signal given that channel inventory build has been a recurring concern across the broader semiconductor industry. Investors will watch the Q3 figure closely as the AI capex cycle enters what some analysts characterise as its first real test of demand sustainability at scale.