AI Infrastructure Alpha

NVDA
AVGO
MU
+6
9 assetsmoderate risk4h

Bets on AI chip companies during rallies, sits in cash and bonds during downturns.

byAmaltash Advisors LLC
Risk level
Moderate
Amaltash classification
Minimum
$1,000
to deploy
Licensing
None
no licensing fee
Advisory fee
up to 1.00%
a year

The idea

Bets on AI chip companies during rallies, sits in cash and bonds during downturns.

Artificial intelligence is the largest capital expenditure cycle in technology history — hyperscalers are spending over $200 billion annually building GPU clusters, and every dollar flows through the same bottleneck: the chips that train the models, the memory that feeds them, the equipment that fabricates both, and the software that turns raw compute into actionable intelligence. This isn't a bet on which AI app wins — it's a bet on the picks and shovels that every AI application must buy regardless of whether the winner is OpenAI, Google, or a startup that doesn't exist yet. The AI infrastructure buildout is still in early innings: enterprise AI adoption is under 10%, sovereign AI programs are launching across 30+ countries, and the transition from training to inference is creating a second wave of chip demand that could exceed the first. But AI stocks don't go up in a straight line — they correct 20-30% between legs higher. This strategy captures the uptrends with conviction and rotates into gold and Treasury bills during corrections, so your capital is either compounding in the strongest AI momentum or earning yield on the sideline. It never sits in a drawdown hoping for recovery — it steps aside and waits for the trend to prove itself again.

Design

How it works

  1. Buys AI infrastructure stocks when momentum turns positive

  2. Sells and moves to safe assets (gold, Treasury bills) when momentum fades

  3. Waits on the sidelines instead of holding through big drops

  4. Focuses on companies that supply the hardware AI needs

Record

How long this strategy has existed

  1. May 20, 2026

    Strategy created

  2. May 20, 2026

    Version 6 created

This strategy has no live track record of its own yet. Results from accounts that deploy it are not combined into a single history.

Performance

Backtested results are for signed-in investors

We don’t publish hypothetical performance on public pages. Sign in to see this strategy’s backtest after the advisory fee and next to the S&P 500, with the assumptions and risks behind it. Backtested results are not returns any account earned.

Sign in to see the backtest

What it holds

9 assets

SEMICONDUCTORS & RELATED DEVICES: 67%COMMODITY CONTRACTS BROKERS & DEALERS: 12%SERVICES-PREPACKAGED SOFTWARE: 10%OPTICAL INSTRUMENTS & LENSES: 7%Unclassified: 4%9assets

Asset mix by sector

  • SEMICONDUCTORS & RELATED DEVICES67%
  • COMMODITY CONTRACTS BROKERS & DEALERS12%
  • SERVICES-PREPACKAGED SOFTWARE10%
  • OPTICAL INSTRUMENTS & LENSES7%
  • Unclassified4%

Target weights for version 6. Actual positions vary as the strategy trades and are not shown here.

The Simulation Lab

Ten bad markets, before your money is in one.

Almost anything looks good in a rising market. That is not a test. So we put the strategies listed here through the same ten bad markets, and this one has been through them. They are stretches of real market history, with names and dates: the 2022 slump, the COVID crash, a year that went nowhere. We picked them before this strategy existed.

Two of them are below. Each one shows our drawing of what that market looked like, and the rules this strategy follows when a market like it turns up. None of this is a score. What actually happened to the strategy is in your dashboard, not on this page.

10bad markets, picked before this strategy existed
8 of 10can run on the real trading days of the time they name
2 of 10we model ourselves, and we mark them as modelled
5 Sep 2026when we last ran this strategy through them

When markets fall

The COVID Crash

A third of the market’s value gone in a month — most of it overnight.

our model of the S&P 500 around Feb–Mar 2020falls about 34% at its worst, then ends about 14% below where it begana shape, not a scale · not this strategy

Index
S&P 500
The real dates
19 Feb – 23 Mar 2020
What the index did
fell 33.9%
How we run it
Real trading days

The fastest fall of that size on record: 23 trading days from an all-time high to the bottom. Much of it happened overnight — big companies opened 8-12% below where they had closed the day before.

The rules this strategy follows here

  1. 1
    What it holds

    9 companies.

  2. 2
    When it sells

    One of them drops 6% to 18% below the price it was bought at.

  3. 3
    What it does

    It sells that one. The rest carry on.

On nine days here, the market opened far below where it closed the day before.

An order set to sell at 6% to 18% down can end up selling far below that. If the market opens below that price, there is no chance to sell at it. That is what this market is here to show, and why we keep it in the set.

When markets go nowhere

The Magnificent Seven Year

A handful of big names carried 2023; most stocks didn’t.

our model of the S&P 500 vs its equal-weighted twin around 2023ends about 8% above where it begana shape, not a scale · not this strategy

Index
S&P 500 vs its equal-weighted twin
The real dates
3 Jan – 29 Dec 2023
What the index did
rose 24.2%
How we run it
Real trading days

The cap-weighted index rose 24% while the equal-weighted version managed under 12%: a handful of very large names carried almost the whole gain and the median stock did little.

The rules this strategy follows here

  1. 1
    What it holds

    9 companies.

  2. 2
    When it sells

    One of them drops 6% to 18% below the price it was bought at.

  3. 3
    What it does

    It sells that one. The rest carry on.

No overnight drops in this market.

A market that ends where it started still charges you for every trade made inside it. This is where the cost of trading adds up fastest. That is why it sits in the set right next to the crash.

When markets rise

The Bull Run of 2016–17

Two calm years when the market just kept climbing.

our model of the S&P 500 around 2016–17ends about 42% above where it begana shape, not a scale · not this strategy

Index
S&P 500
The real dates
4 Jan 2016 – 29 Dec 2017
What the index did
rose 30.8%

When markets go nowhere

When Tech Handed Over to Oil

Yesterday’s winners became 2022’s losers.

our model of the S&P 500 around 2021–22rises about 27% at its turn, then ends about 2% above where it begana shape, not a scale · not this strategy

Index
S&P 500
The real dates
4 Jan 2021 – 30 Dec 2022
What the index did
2.2% higher

Sign in to see how this strategy did in each one.

The results open in your dashboard, where we can explain what they mean for you. They are simulated results, so we never show them here.

Open the stress tests

trade.amaltash.com/marketplace/ai-infrastructure-alpha?tab=stress

All ten markets

  • The Bull Run of 2016–17When markets rise · modelled · shown above Tested
  • The 2020 ReboundWhen markets rise Tested
  • The 2022 SlumpWhen markets fall Tested
  • The COVID CrashWhen markets fall · shown above Tested
  • The Christmas 2018 ScareWhen markets fall Tested
  • The Year That Went NowhereWhen markets go nowhere · modelled Tested
  • The Wild Swings of Late 2022When markets go nowhere Tested
  • The Quiet YearWhen markets rise Tested
  • The Magnificent Seven YearWhen markets go nowhere · shown above Tested
  • When Tech Handed Over to OilWhen markets go nowhere · shown above Tested

These ten are the whole set. We picked them for the damage they did — the fastest crash on record, a year that went nowhere, a grind that punished every rally — not for how they make anything look.

8 of these 10 can run on the real trading days of the time they name. We model the other 2 — The Bull Run of 2016–17 and The Year That Went Nowhere. A strategy needs a stretch of history to warm up on before a test starts, and our price data does not go back far enough to give these that. So we built stand-ins that behave like those years, rather than replays of them. A modelled market is not a forecast, and it is not what would have happened. The shapes drawn above are our models of those markets — never this strategy — and they have no scale. The dates and index moves next to each one are the real figures for the period it is modelled on. Nothing on this page says how any strategy did in these tests.

Fees & minimums

What it costs

Licensing fee, once
None
Advisory fee, yearly
up to 1.00%
Minimum to deploy
$1,000

What $10,000 costs, at a flat balance

1 year

$100

3 years

$300

5 years

$500

The maximum advisory fee. Your actual fee rises and falls with your balance. Other costs, such as spreads and regulatory fees, may apply.

Documents

Risk Disclosure: Trading in financial instruments involves substantial risk, including the possible loss of your entire investment, and may not be suitable for all investors. Prices can be affected by external factors such as financial, regulatory, or political events. Trading on margin or with leverage increases potential losses. Past performance is not indicative of future results.

Not Financial Advice: The information provided on this platform is for informational purposes only and does not constitute investment, financial, or trading advice. We do not recommend any particular trading strategy or instrument. Please conduct your own research and consult with a qualified financial advisor before making investment decisions.

Get Started

Ready to Deploy This Strategy?

Connect your brokerage account and turn on the strategy. We then place trades that follow its rules and your limits. No coding needed.

Start Trading Now →

No licensing fee

+ up to 1.00%/yr

Sign in to deploy