TrendArc AI 100 Index

What it measures. The AI 100 against the funds you could actually have bought instead (QQQ, SPY and the equal-weight S&P, RSP), against gold, and against the M2 money supply. All rebased to 100 and total-return. How to read it. Log scale: a straight line is steady compounding, a steepening line is acceleration. The gap between the magenta line and the rest is the return you gave up by owning the benchmark instead of the supply chain. Gold and M2 are the debasement yardsticks: gold is what money runs to, M2 is how much of it there is, so anything below the M2 line has not really grown at all once the unit of account is accounted for. Why it matters. Benchmarks are backward-looking by design. They tell you who won the last cycle, not who receives the marginal dollar in the next one. The software era rewarded margin capture; the AI era rewards capital expenditure, and the receivers of that spend are the picks and shovels. This chart is the evidence for that thesis, and the reason we track the theme on its own index rather than on the Nasdaq. Daily-rebalanced equal weight; an index, not an investable product.

What it measures. How far the index has travelled over the last 50 trading days, the speed of the move rather than its direction. How to read it. The +20% line is the crowding threshold: above it, the theme is absorbing money faster than fundamentals change, and positioning is doing the work. The −10% line is washout: capitulation, historically where the theme has paid to add rather than trim. What matters most is not the level but the roll-over, a peak above +20 that turns down while breadth thins is the speed-crash setup. Why it matters. Speed crashes are a positioning event, not an earnings event. They are invisible in valuation and invisible in the headline index; they show up first as velocity. This is the fastest of our three clocks and usually the first to move.

What it measures. How many names are actually participating. Fast is the share above their 20-day average (the short-term pulse); slow is the share above their 200-day (the structural trend). How to read it. Above 85% is overheated, nearly everything is extended at once, which is a condition, not yet a signal. Below 30% is washed out. The warning is divergence: the index holding near its highs while fast breadth rolls over means a shrinking group of names is carrying the theme. Why it matters. This is the confirmation clock. Crowding tells you the trade is full; breadth tells you it is starting to empty. Together they are the setup we act on, and because this index is equal-weight, thinning shows up here that a cap-weighted benchmark would mask entirely.

What it measures. Three volatilities on the same 20 to 30 day horizon. AI 100 realized is how violently the theme itself is actually moving. Factor volatility is the same measure across momentum, value, quality and high-beta positioning, so it reads the style complex rather than the theme. Implied is what the options market is charging for the next month. How to read it. The theme's own line sits structurally above the other two: an equal-weight basket of high-beta supply-chain names is simply a more volatile thing than the broad market, so watch its direction, not its level. On the other two the signal is the crossover: in calm regimes implied sits above realized, because the market pays for protection it does not end up needing. When factor volatility pulls above implied, positioning is being repriced rather than the market level. Why it matters. This is the clock that catches damage the price line hides. Crowded factors tend to unwind before the theme does, because the unwind starts inside the crowd and not at the headline, so factor stress rising while the index is still calm is a warning about what comes next.

How the signals are generated. Exhaustion first: we run ArcCount on the index's own weekly and monthly bars, exactly as on any stock card. A single completed sell-13 is a useful mark and is pinned in purple. A double, where the countdown-13 and the strict-13 complete on the same bar, is the highest-conviction exhaustion signal in the arc guide, so weekly and monthly doubles are promoted to full red SELL pins in their own right. The model adds one more: a weighted composite that fires when an ArcCount 13 lands while the 50-day rate of change is above +20% and breadth is already dropping off its recent peak. The crowding gate is the whole trick: exhaustion in a calm tape is noise, exhaustion in a crowded one is a speed crash forming. Buys are the mirror image: the green BUY needs the rate of change to have washed out below −10% and then turn back up, with breadth lifting off the floor and volatility rolling over, capitulation that has stopped getting worse, not a falling knife. Why it matters. Speed crashes are positioning events. They do not announce themselves in earnings, and the headline index hides them while a few mega-caps hold it up. These pins are the moments the theme's own internals said the crowd was full. How to read it. A red mark identifies the crowded end of the theme, the point at which positioning is stretched rather than fundamentals broken. What is crowded and what is not tend to separate from here; historically the uncrowded end has included defensives, precious metals and cash. A green mark identifies the other end, where the rate of change has normalised and volatility has come back in. The distance between the two is where the theme’s own internals do most of their work. Research and monitoring, not investment advice.

What it measures. The AI 100 broken into the 14 functions it is built from, each one an equal-weight index of its own members. Every line starts at 100 on the first day of whichever range is selected, so All measures from January 2016 and 1Y measures from a year ago. Changing the range re-bases the chart rather than zooming into it, and that is the point: the leaders and laggards of a window are only legible when the lines all set out from the same place, instead of entering it wherever a decade of divergence had already left them. The left scale is the index level, the right scale is the same thing read as cumulative gain since the start of the range, and the scale is logarithmic, so equal vertical distances are equal percentage moves. The multiple beside each name in the legend follows the range too. How it is built. Money in at the start, split evenly across the vertical's members, then held. "The start" means January 2016 in every range: a shorter window re-scales that same basket's history so it opens at 100, it does not re-form the basket at equal weights on that day. So 1Y reads as what the 2016 vintage of the vertical did over the last year, drifted weights and all. A member that listed later (ARM, GEV, Palantir) joins on its first traded day at the basket's average position size, funded from the existing holdings, so an addition moves the line's future, never its level. Note this differs from the headline AI 100, which rebalances back to equal weight daily. The verticals do not, deliberately: daily rebalancing pays a volatility bonus that grows as the basket shrinks, which would have handed the two-name Memory vertical a large advantage over the seventeen-name Materials vertical for reasons that have nothing to do with the AI build-out. How to read it. The ranking is the point. The spread between the top and the bottom line is the cost of being right about the theme and wrong about where in the chain to stand: every one of these companies receives AI capital expenditure, and they are still separated by more than an order of magnitude. Watch where the lines change order, because that is the bottleneck moving. Early in the cycle the scarce thing was logic and the design tools; as the build-out turned physical the scarcity moved toward memory, the equipment that makes it, then power, cooling and the grid. A vertical that steepens while the others flatten is where the marginal dollar is going now. Why it matters. The index tells you whether to be in the theme. This chart tells you where in the theme, which is the larger decision. It is also the honest risk view: the leaders here are the crowded end, so when the model turns defensive, this is the picture of what has the most to give back. Daily-rebalanced equal weight for the reference line, buy-and-hold equal weight for the verticals. Indices, not investable products. Backup Energy & Storage starts in 2018, the first day any of its members had listed.