Nicholas Mersch’s Top Picks for May 13, 2026

Nicholas Mersch, Portfolio Manager, Purpose Investments

Focus: Technology stocks

Top Picks: Micron, Lumentum, Vicor

MARKET OUTLOOK:

We are still structurally underbuilt AI infrastructure.

This showed up in a big way in a certain pocket of the market (semiconductors) as more fuel was added to the fire when two key things happened over earnings period:

1. Capex numbers got bumped up again.
2. Hardware now has all the pricing power.

There’s a very interesting thing going on in the supply chain here, but you need to zoom out before zooming in to understand what is driving this supply and demand imbalance.

Invest in the companies cashing the cheques, not the ones writing them.

It is is too early to look at software, and its not too late to look at semis

The value chain is evolving in the following manner:

Each step function of artificial intelligence (AI) requires more compute: Each evolution of AI (generative, reasoning and agentic) sees token costs declining about a hundred times at the same time usage increases ten thousand, driving more demand for more compute and more dollars spent.

Everyone is fighting for this scare resource: Token explosion and service outages are results of massive demand for AI, driving companies to fight for scarce capacity.

Consensus has been wrong on Capex: This usage eruption is leading to Capex spiking across the board, way outpacing prior estimates as the goalposts continue to shift.

Cash flow is flowing from megacaps to semis, plus infrastructure: This is driving pricing power for semiconductor and infrastructure companies. Stocks are up on a stick, but earnings revisions are outpacing stock price performance.

We are seeing more return on AI: Demand is further validated by AI startup revenue run ramp, and from growing backlog from hyperscalers. However, a large portion of backlog from hyperscalers is the AI startups, creating some concentration/circularity risk.

Stay long semis, stay long infrastructure, stay away from software.

• Market-moving news, fast: Get the BNN Bloomberg App now
• Sign up for the Market Call Top Picks newsletter at bnnbloomberg.ca/subscribe

TOP PICKS:

Micron (MU NASDAQ)

Micron is the scarcity trade inside the scarcity trade. High-bandwidth memory capacity is effectively sold out through 2026, high bandwidth memory 4 (HBM4) is sampling to customers, and dynamic random-access memory (DRAM) pricing keeps drifting higher on supply discipline from the three remaining global producers. The single most important corporate signal of the quarter was Meta hiking capex to US$125 to US$145 billion dollars and naming memory pricing as the reason. That is the entire Micron thesis in one sentence, delivered by a customer, unprompted. The setup is allocation-driven, not share-driven. SK Hynix is the high bandwidth memory (HBM) leader, Micron is the credible number two, and Samsung is still struggling on qualifications. That qualification gap is the margin opportunity, and it is the one bears need to watch. The day Samsung clears Nvidia HBM qualification at scale, the pricing premium gets diluted. Until then, Micron prints. Free cash flow has inflected materially off the trough. Idaho fabrication investment provides geopolitical hedging and CHIPS Act tailwind. Forward earnings look cheap if the cycle sustains and dangerous if it turns, because memory always turns eventually. The cycle question is how many more quarters you get. The bull view says HBM intensity per AI system is structurally higher than legacy compute, so this cycle runs longer than skeptics expect. The bear view says memory always over-earns then under-earns, and HBM capacity expansions in 2026 and 2027 will eventually overshoot demand. Cyclical watch-list, not a forever name. Trade the cycle.

Lumentum (LITE NASDAQ)

Lumentum makes the lasers and optical components that move data between computer chips inside AI data centres. When tens of thousands of Nvidia graphics processing units (GPUs) need to communicate with each other, copper wires cannot carry that much information that fast, so the signal travels over fiber optic cable using tiny lasers. Lumentum makes those lasers. The most recent quarter showed revenue of US$808 million, up 90 per cent from a year ago, and management guided next quarter to roughly US$1 billion with operating profit margins reaching 32 per cent. The company has said openly that it cannot manufacture enough product to satisfy demand and is shipping about 30 percent less than customers want to buy. Nvidia stepped in with a US$2 billion equity investment and a multibillion-dollar order commitment so Lumentum can build new factories. Three products drive the story over the next two years. The first is faster optical transceivers running at 1.6 terabits per second, which become standard on the next generation of AI servers. The second is optical circuit switches, a newer product that replaces traditional electrical switches and saves significant energy inside data centres. The third is co-packaged optics, which places the lasers directly onto the networking chip itself, an architecture every hyperscaler is moving toward. We own Lumentum because it sits in the narrowest part of the AI supply chain. Whoever wins the GPU race needs more of what Lumentum makes, and the company cannot expand fast enough to satisfy customers.

Vicor (VICR NASDAQ)

Vicor makes small power-conversion modules that sit on the same circuit board as Nvidia’s GPUs and feed them clean electricity. Modern AI chips draw enormous amounts of power, well over 1,000 watts per processor, and the larger the chip the harder it becomes to deliver that current without wasting energy as heat. Vicor’s solution is called Vertical Power Delivery, and it stacks the power module directly underneath the chip instead of placing it off to the side. That layout is more efficient, uses less board space, and runs cooler than competing approaches. The latest quarter brought US$113 million in revenue, up 20 per cent year over year, with gross margins expanding roughly 800 basis points to 55 per cent. The number worth focusing on is backlog. Customer orders waiting to ship jumped 70 per cent in a single quarter to US$301 million, and new orders are arriving at more than twice the rate of shipments. Management is guiding to roughly US$570 million in 2026 revenue with a longer-term target around a US$1.5 billion annual run rate as the Massachusetts factory expands. A hidden royalty business adds optionality. Vicor holds patents on core power-conversion architectures and is now collecting licensing fees from competitors who use them, with management projecting hundreds of millions over time. The company is small, the stock is volatile, and execution on factory expansion is the primary risk. We own Vicor because the AI power problem worsens every year and Vicor has the best technical answer at the chip-package level.

PAST PICKS: JAN. 19, 2026

Taiwan Semiconductor (TSM NYSE)

Then: US$342.40

Now: US$396.02

Return: 16%

Total Return: 16%

Emcor (EME NYSE)

Then: US$698.69

Now: US$919.85

Return: 32%

Total Return: 32%

Fluence (FLNC NASDAQ)

Then: US$27.08

Now: US$22.19

Return: -18%

Total Return: -18%

Total Return Average: 10%