AI Infrastructure Stack 6
Calibrated launch value from the rulebook. This basket carries no order panel today.
Why this basket exists
What the construction is trying to capture, who it is built for, and how it is meant to be sized inside a wider allocation.
One layer beneath NVIDIA. Applied Materials and Lam Research sell the equipment that lets foundries make advanced chips. Arista Networks runs data-center networking at the hyperscalers. Marvell ships custom silicon for ASIC-bound workloads. Super Micro builds the servers Big Tech actually rolls into racks. Micron supplies the HBM memory that bottlenecks AI training. Higher beta than OSEM7, lower market cap per name.
- Equities family · Growth risk tier
- Rule-based, divisor-method NAV
No committee note is published for this basket yet, so the catalogue description above stands in until one is written.
6 holdings, drawn to scale
Each tile's area is its weight in the basket, so the picture and the table below it are the same data read two ways.
| Holding | Share | Weight | |
|---|---|---|---|
| 01AMATon | 22% | ||
| 02LRCXon | 20% | ||
| 03ANETon | 18% | ||
| 04MRVLon | 14% | ||
| 05SMCIon | 14% | ||
| 06MUon | 12% | ||
| Total | 100% |
Why each name earned its seat
Research notes per constituent: the case for the asset, the regime that suits it, and the risk the seat carries.
- AMATon22%Infrastructure
Applied Materials is the largest semiconductor-equipment maker by revenue. Provides the deposition and etch equipment for leading-edge logic and DRAM.
Performs in capex-cycle ups. Sensitive to fab-capex deferrals and China-export-control headlines.
capex-cycle exposure - LRCXon20%Growth
Research note pending for this holding. Its weight and its role in the mandate are published above.
- ANETon18%Infrastructure
Arista Networks is the dominant data-center networking vendor for hyperscalers. AI-cluster networking demand drives multi-year revenue growth.
Performs in hyperscaler-capex cycles. Sensitive to hyperscaler-spending-pause headlines.
hyperscaler concentration - MRVLon14%Growth
Research note pending for this holding. Its weight and its role in the mandate are published above.
- SMCIon14%Growth
Research note pending for this holding. Its weight and its role in the mandate are published above.
- MUon12%Growth
Micron is the largest US memory and storage chip manufacturer. Supplies HBM that bottlenecks NVIDIA and AMD AI accelerator production.
Performs in memory-cycle ups and AI-capex peaks. Sensitive to memory-pricing surprises and inventory corrections.
memory-cycle exposure
What the numbers say
Simulated on historical constituent prices. A Sharpe is only published once the measured window clears 90 days, and drawdowns keep their sign.
Every figure above is simulated on historical constituent prices for a basket that has not traded. It describes how the rules would have behaved, not how they will behave. OAIE6 is part of a public preview.
Max expected loss over the horizon · 95% confidence · historical distribution. There is a 5% chance the actual loss is larger.
| Window | Return | Sharpe | vs BTC |
|---|---|---|---|
| Full window | +832.6% | 1.49 | +1.04 |
| 1 year | +184.4% | 4.06 | +4.84 |
| 6 months | +72.6% | 3.67 | +2.15 |
| 3 months | +11.6% | 0.88 | −0.32 |
| 1 month | +29.3% | Not graded | · |
Returns are simulated over each window. A Sharpe is published only once that window clears 90 days, so the shorter rows read Not graded and carry no comparison with bitcoin. Short windows flatter and punish at random; read the full window first and treat the rest as texture.
What could go wrong
Priced off this basket's own weights rather than off a generic disclaimer, so every figure below can be checked against the table above.
Higher beta than the broad market. Drawdowns of 40 to 60% during risk-off regimes are within historical range. Size accordingly.
- Top-name weightLow
AMATon is the largest position at 22%. Capped by methodology but still drives index direction.
- Short price historyHigh
AMATon, LRCXon, ANETon and 3 more have under a year of usable history. Backtest metrics carry wider error bars.
- Routing dependenceMedium
Routes through Dinari (stocks) and Backed Finance (ETFs). Settles on the EVM rail. KYC at the OLTA layer, not available in the US today.
| Scenario | Loss on net asset value |
|---|---|
| AMATon loses half its valueLargest holding, 22.0% of the basket | −11.0%Mechanical |
| The three largest holdings fall 30%AMATon, LRCXon, ANETon · 60.0% of the basket together | −18.0%Mechanical |
| The deepest simulated fall repeatsRecorded once in the window, 291 days to recover | −43.2%Recorded |
| A bad month, one in twenty95% historical value at risk over 30 days | −14.4%Measured |
The mechanical rows hold every other holding still, which is the optimistic case. This basket measured 0.29 correlation to bitcoin over its window, which is loose enough that the holdings can move apart. Even so, a shock rarely lands on one name alone. Read the figures as a floor on the damage, never a ceiling.
The research behind the construction
Tokenized equity execution is not open yet
Single stocks route through Dinari and ETF coverage through Backed Finance, settling on the EVM rail once identity checks clear. The execution surface for this family arrives with the next release. Composition, weights and the measured record above are current.