Nick Mersch’s Top Picks for Aug. 20, 2026

Nick Mersch, Portfolio Manager, Purpose Investments

Focus: Technology stocks

Top Picks: Palantir, Snowflake, DigitalOcean

MARKET OUTLOOK:

The market is telling us we still need more compute. The goalposts shifted again over the earnings period, where we’ll spend approximately US$1 trillion on artificial intelligence (AI) Capex this year, and approximately US$1.3 trillion (up from approximately US$1.1 trillion estimates after first quarter prints). What we are seeing here is that the goalposts keep moving out. All of the free cash flow is going from hyperscalers to semiconductors, but I think that hyperscalers will monetize this spend faster than people realize.

Hold semi’s but time to be selective. We are seeing unprecedented earnings growth in the semiconductor cohort. Micron’s earnings per share (EPS) is going from US$7 in 2025 to US$150 in 2027. Sandisk’s is going from US$70 to US$250. So we’re seeing step-function cash creation here. There is no question on valuation - the question is on peak earnings. While some argue the demand is now permanent and structural, this is still a cyclical industry. I like it because they now have longer term contracts, but this doesn’t negate the fact that we’ll have more capacity come, as the margins are just too strong right now. Hold them for a couple more quarters, then look to see any weakness in margin for an indicator of the cycle turning.

It’s OK to buy software again - but not just any software…infrastructure software. I think its ok to buy software in the token path. By that I mean, any software that enables AI agents. Observability and security software stocks never got caught in the SaaS-pocalypse, and rightfully so. But I think you can start to add to names in the space like Snowflake and Palantir now that we can see how the data estate needs to be in order to optimize the business use-cases for these systems.

Time to add to megacap. Despite spending massive amounts on Capex, we are finally starting to see some return for this. The cloud players put up massive growth on a massive base as they continue to effectively monetize the AI giants (OpenAI and Anthropic). If we look at the useful life of graphics processing units (GPUs), Nebius showed us that they signed a contract for an A100 out to 2029. As a reminder, the A100 came out in 2020. So depreciation lifetime actually longer. Megacaps own all the largest data centers that are monetizing these chips. Return on Investment (ROI)-a-plenty. They are also eating their own cooking. Revenue per employee figures are through the roof, showing tons of operating leverage

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

TOP PICKS:

Palantir (PLTR NASD)

Palantir has become the cleanest public expression of enterprise AI software demand, converting foundation models into deployed workflows through Gotham, Foundry and AIP.

The Aug. 3 print was the proof point: revenue of US$1.94 billion grew 93 per cent year over year, with U.S. commercial up 149 per cent to US$764 million and U.S. government up 90 per cent to US$809 million, and generally accepted accounting principles (GAAP) net income came in at US$1.07 billion.

The adjusted operating margin hit 62 per cent and the Rule of 40 score reached 155, a combination of hypergrowth and profitability with no precedent at this scale. Management raised full-year revenue guidance to US$8.15 billion, implying 82 per cent growth, and guided adjusted free cash flow to US$4.5 to US$4.7 billion, while remaining US commercial deal value more than doubled to US$6.24 billion. Karp framed the quarter around unleashed demand for AI sovereignty, and the entrenched government franchise gives that framing real substance as allied nations stand up sovereign stacks. The debate is entirely about price.

The stock had lost roughly 29 per cent year to date heading into the print as the AI software trade deleveraged, then jumped 12 per cent on the results, and it still trades north of 100 times earnings. At that level the stock is a duration bet on hypergrowth persisting for years, which means any deceleration in US commercial bookings would get amplified violently through the multiple. The fundamentals are currently unimpeachable, and the margin of safety embedded in the price is approximately zero.

Snowflake (SNOW NYSE)

Snowflake has repositioned from cloud data warehouse to AI Data Cloud, and the consumption model is finally showing AI leverage. Q1 fiscal 2027 delivered product revenue of US$1.334 billion, up 34 per cent and marking a third consecutive quarter of acceleration from 26 per cent a year ago through 30 per cent last quarter, the strongest sequential dollar growth in company history, with net revenue retention climbing to 126 per cent.

Remaining performance obligations grew 38 per cent to US$9.21 billion and the million-dollar customer count reached 779, with 46 crossing the threshold in the quarter against 26 a year earlier. Non-GAAP EPS of US$0.39 beat the US$0.32 consensus on an 11.9 per cent operating margin, though the company still ran a GAAP loss of roughly US$310 million given heavy stock compensation. Management raised full-year product revenue guidance to US$5.84 billion, or 31 per cent growth, up from a prior 27 per cent, and paired the print with a US$6 billion five-year AWS commitment that leans into Graviton and GPU capacity. The strategic pitch under Ramaswamy is the agentic enterprise, with Snowflake Intelligence giving business users natural language access to governed data while builders assemble agents and pipelines directly on the platform.

Consumption pricing means AI workloads land directly in revenue, which cuts both ways in a downturn. Databricks remains the structural overhang, now carrying a US$190 billion private valuation. The next catalyst is the Q2 print, guided to US$1.415 to US$1.420 billion in product revenue, or 30 per cent growth.

DigitalOcean (DOCN NYSE)

DigitalOcean has re-rated from sleepy Server Message Block (SMB) developer cloud into an AI inference story under Paddy Srinivasan, who has repositioned the company as an AI-Native Cloud purpose-built for inference and agentic workloads.

The Aug. 4 print validated the pivot: revenue of US$281 million grew 29 per cent year over year, more than double the growth rate a year ago, with AI customer ARR up 212 per cent to US$234 million, million-dollar-plus customer ARR up 214 per cent to US$259 million, and a record US$93 million of incremental ARR. Inference services grew 800 per cent and now represent over 70 per cent of AI customer ARR, with the inference engine launched in late April already counting more than 6,000 customers. RPO expanded more than tenfold year over year as the company signed multiple nine-figure annual commitments, with weighted average contract life stretching from 1.6 years to over three, landing at US$894 million.

Management raised the full-year outlook to US$1.17 to US$1.18 billion, implying 30 to 31 per cent growth, and projected roughly 35 per cent exit growth by Q4 with confidence in 50 per cent-plus growth for 2027. The tensions sit underneath the acceleration: adjusted EBITDA held a 40 per cent margin but GAAP operating income fell 18 per cent and cash flow margins narrowed, equipment financing and lease obligations have climbed substantially, and new data center capacity arrives only in 2027 and 2028, which caps near-term supply while concentrating the book in a handful of large AI accounts whose usage now drives companywide results. The long-tail Droplet base is becoming ballast for a leveraged inference bet, and counterparty quality on those nine-figure commitments is the question worth pressing.

PAST PICKS: MAY 13, 2026

Micron (MU NASD)

Then: US$803.63

Now: US$937.10

Return: 17%

Total Return: 17%

Lumentum (LITE NASD)

Then: US$1030.37

Now: US$827.60

Return: -20%

Total Return: -20%

VICOR (VICR NASD)

Then: US$309.27

Now: US$217.35

Return: -30%

Total Return: -30%

Total Return Average: -11%