Risk framework
Core positions should be sized at 5-10% for true structural themes. Tactical trades should stay in the 1-5% range for event-driven or momentum setups. Speculative names should be limited to less than 1-2% because the downside can be permanent.

 TTD
The cleanest event-driven long on the board is The Trade Desk. CEO Jeff Green bought roughly $148 million of stock in the open market, which is a very rare level of conviction and helped trigger a sharp move higher. The market is also starting to price in a possible strategic relationship with OpenAI, which makes the stock look like a momentum-plus-catalyst setup rather than a simple insider buy.

Trade setup: bullish bias, buy pullbacks after the initial move, and keep the position tactical at 1-5%. The trade is invalidated if partnership rumors are denied or if the broader ad-tech environment weakens.

 NVDA and AMD
Nvidia and AMD now face a direct margin issue from the 15% revenue share tied to AI chip exports to China. This is not just a one-time cost; it changes the economics of selling into China and creates a structural headwind to gross margin. If the market starts to view this as a permanent feature, the valuation multiple should compress further.

Trade setup: bearish to underweight, ideally sold into strength rather than chased on weakness. Keep the size tactical at 1-5%. The bearish case weakens only if companies fully pass the cost to end customers or if the impact is priced in much faster than expected.

 ASML
ASML is now facing a more explicit policy overhang through the MATCH Act and FDPR-related restrictions. The main issue is not only new equipment sales to China, but also servicing and support, which are important for recurring revenue and long-term customer relationships. That means the China exposure increasingly deserves a political discount.

Trade setup: avoid or short on strength, especially if the stock keeps failing at resistance. This should remain a tactical trade with 1-5% sizing. The key invalidation is a dilution of the bill or a meaningful extension of deadlines.

 MU and EWY
Micron is the strongest core long in the group. Appaloosa reportedly increased its Micron stake sharply and also opened a large position in South Korea via EWY, which looks like a broader bet on the memory stack and the Korean semiconductor ecosystem. The important point is that capital is rotating from logic toward memory, and HBM appears to be the next bottleneck in the AI supply chain.

Trade setup: long, buy into trend continuation and accumulate on weakness. This deserves core sizing at 5-10%. The main risk is a slowdown in AI capex or an eventual HBM supply glut.

 Defense and AI procurement
Defense remains a structural long theme. Denmark’s fast-tracked Lockheed Martin contract shows that sovereign buyers are increasingly willing to bypass normal procurement channels when national security is at stake. At the same time, Pentagon spending on autonomous AI coding suggests the market is moving away from manual labor models and toward software-defined defense capabilities.

Trade setup: long the sector, preferably through breakouts in defense ETFs or selective names with direct exposure to the modernization cycle. This belongs in core allocation at 5-10%. The thesis weakens if budgets are cut or if procurement returns to slower peacetime patterns.

 BR
Broadridge is a quieter but more defensive insider signal. The CEO bought more than $1 million worth of stock, which usually matters more when the business is stable, recurring, and not highly cyclical. In that kind of setup, insider buying often signals confidence in margin durability rather than a short-term trading catalyst.

Trade setup: long on pullbacks, core-sized at 5-10%. The trade is invalidated by a meaningful earnings miss or a disruption in its core processing franchise.

 Speculative names
BMNR and RGTI sit in the speculative bucket. These are the kind of names that can work only if you treat them like venture-style public market options, not like normal equity positions. The right way to approach them is with very small sizing and a willingness to lose most or all of the capital allocated.

Trade setup: long only with strict risk limits, ideally below 1-2% portfolio weight. The thesis breaks if commercialization takes too long or if liquidity deteriorates before the technology matures.

 Final portfolio view
The portfolio should be structured as a barbell. On one side, hold core longs in MU and defense-related names. On the other side, use tactical longs in TTD and tactical shorts in NVDA, AMD, and ASML. The main macro idea is that the market is underpricing both the rotation inside the AI stack and the political risk embedded in China-exposed semiconductor names.