2 AI Stocks to Avoid (Including BigBear.ai) and 1 to Buy Now

The article is bearish on BigBear.ai and C3.ai, citing deteriorating or weak fundamentals: BigBear.ai revenue fell from $146 million to $128 million from 2021 to 2025, while C3.ai's revenue is expected to decline from $389 million in fiscal 2025 to $251 million by fiscal 2028. By contrast, Broadcom is highlighted as the preferred AI winner, with AI chip sales up 65% to $20 billion in fiscal 2025 and projected to reach $60 billion-$90 billion by fiscal 2027. Overall, the piece argues that Broadcom remains a best-in-breed AI chip play while the two smaller AI software names face competition and self-inflicted growth issues.

Analysis

The market is repricing AI spend from “model narrative” to “economic moat,” and that’s a bad backdrop for subscale software wrappers with weak retention and high customer concentration. BBAI and AI both face a second-order problem: as hyperscalers and large enterprises standardize on vendor-native tooling, point solutions lose pricing power and get forced into services-heavy mixes that compress margin before revenue growth can reaccelerate. In other words, the issue is not just slower top-line growth; it is a deterioration in the quality of revenue that tends to show up one or two quarters before headline estimates are cut.

Broadcom’s edge is less about generic AI demand and more about budget migration inside hyperscalers from experimentation to deployment. Custom silicon wins when customers have already proven workload scale, because the ROI is measured in power, capex efficiency, and supplier leverage versus Nvidia—not in model performance alone. That creates a durable “picks-and-shovels-to-the-picks-and-shovels” effect: every additional AI inference workload that becomes productionized increases ASIC relevance, while also making Broadcom’s networking and software attach harder to displace.

The contrarian read is that the bearish cases on BBAI and AI may still be too conservative on downside timing, but not necessarily on ultimate fundamentals. These names can stay expensive for a while if retail sentiment and government contract headlines offset deteriorating fundamentals; the catalyst for a true de-rate is likely a miss-and-cut cycle, not just slow growth. For AVGO, the obvious risk is valuation compression if AI capex slows, but that risk is more likely to matter over 6-12 months than over the next few weeks, because backlog conversion and hyperscaler deployment cycles should keep estimates moving higher before any macro slowdown shows up.

One subtle winner outside the headline is NVDA: even if custom ASIC share rises, it can be additive rather than purely substitutive if total AI infrastructure spend keeps expanding. GOOGL and META also benefit from Broadcom’s thesis because in-house silicon reduces inference cost, which can extend the economic life of their AI products and support faster monetization. The key cross-asset signal to watch is whether ASIC demand keeps outrunning general AI software adoption; if that spread widens, it argues for long infrastructure, short software beta.

Key Decisions for Investors

• Short BBAI on any post-headline strength; use a 1-3 month horizon and treat 15x sales as vulnerable to multiple compression if growth decelerates further. Best risk/reward is into optimism spikes or contract announcements.
• Short AI against long AVGO as a sector pair trade; thesis is that capital is shifting from low-quality AI application layer to cash-generative infrastructure with clearer ROI. Consider 3-6 month horizon with a stop if AI reacceleration or major operating margin improvement appears.
• Add to AVGO on pullbacks; favor a 6-12 month hold given hyperscaler ASIC adoption is still in early monetization. Upside comes from estimate revisions, while downside is buffered by diversified software cash flows.
• For options, buy AVGO call spreads 6-9 months out rather than outright equity if you want leveraged AI infrastructure exposure with defined risk. The trade benefits from estimate drift and should outperform if AI capex stays resilient.
• Watch NVDA and GOOGL as indirect beneficiaries; if Broadcom’s ASIC narrative strengthens, these names likely remain supported by broader AI spend rather than losing share outright. Use weakness in either as confirmation that the market is over-discounting custom silicon substitution.