Earnings call transcript: Oncopeptides Q2 2026 sales jump 62% as stock rises

Oncopeptides reported Q2 2026 net sales of 31.1m SEK (+62% YoY) and 56.5m SEK for H1 (+74% YoY), with gross margin at 98% in H1. Operating expenses fell 2% to 146.9m SEK in H1, improving the EBIT loss to 86.2m SEK (from 116.1m SEK), and the company ended the quarter with 158m SEK in cash after a March rights issue that raised 167m SEK. Management reiterated ~300m SEK full-year 2026 operating expense guidance and said it remains on track for positive cash flow during 2027, supported by progress including an EMA Type II label-expansion submission and a first patient dosed in its glioblastoma window-of-opportunity study. Shares rose 10.16% to $1.875 on the update.

Analysis

NVDA’s print reinforces that AI infrastructure spend is still not just durable but concentrated in a handful of vendors, which is bad for dispersion: the more budget flows to one platform, the more the rest of the stack must justify its share with superior attach rates. Near term, the main winners are the bottlenecks around the platform — HBM memory, advanced packaging, networking, and power/cooling — because those line items tend to re-rate when GPU demand remains unconstrained. The subtle loser set is the broad semiconductor beta basket: if one name is capturing most of the incremental AI wallet, late-cycle “AI semis” without clear design-win visibility can underperform even in a strong tape.

The bigger risk is not the quarter, it’s the next two catalysts: capex digestion and gross-margin normalization across the ecosystem over 1-3 months. If hyperscalers start signaling that current spend is more deployment-heavy than expansionary, high-multiple names tied to AI buildout can de-rate before revenue rolls over. Over 6-18 months, the question becomes whether inference monetization broadens the customer base enough to sustain this capex curve; if not, NVDA remains excellent operationally but increasingly vulnerable to multiple compression rather than earnings misses.

Consensus may be underestimating how much of the good news is already in the distribution. The strongest contrarian angle is to favor the picks-and-shovels with pricing power but less headline dependency over the obvious GPU proxy. There is also a tactical trade in identifying vendors that are leverageable to AI spend but less protected by supply constraints — they can lag when buyers concentrate orders into the market leader’s ecosystem. For the rest of the data set, SYBT/TGT/TSTS look like noise versus the AI tape and likely should not be forced into the trade framework.

Key Decisions for Investors

• Stay long NVDA on 1-2 week momentum, but tighten risk: use a trailing stop if the stock fails to hold post-earnings gap support; upside is still intact, but the easy multiple expansion may be behind us.
• Pair trade: long AVGO or MRVL vs short a diversified AI semi basket/ETF proxy over 1-3 months to capture wallet-share concentration and relative pricing power.
• Add exposure to the constrained infrastructure beneficiaries (e.g., ANET, SMCI selectively, or HBM-linked names where liquid) for 1-3 months; these are the cleanest second-order winners if AI capex stays elevated.
• Fade overowned “AI beta” names that have run ahead of visible revenue conversion; use calls to express downside convexity if hyperscaler commentary shifts to digestion.
• Alert item: if the next hyperscaler capex guide rolls over or NVDA gives back the post-print gap, reduce sector exposure aggressively — that would be the first credible falsifier of the bullish tape.