PSI ETF at $152.85, Down 19.2% From $189.24, After Returning 102.37% in 2026 Against SMH's 66.2% — AUM Reaches $3.82B

The Invesco Semiconductors ETF traded at $152.85 on Thursday afternoon, down 0.40% or 62 cents from Wednesday's $153.47 close. The session opened at $151.23, ran to $155.40, printed a low of $151.10, and settled back into the middle — a 2.85% intraday range on a day the Nasdaq Composite fell 1.5%.

Market capitalisation stands at $3.82 billion. Average daily volume runs roughly 461,000 shares.

The two numbers that frame everything else are the 52-week extremes: a high of $189.24 and a low of $57.36. At $152.85, PSI sits 19.2% below its peak and 166.5% above its trough.

Both facts are true simultaneously and both matter. This fund returned 102.37% from December 31, 2025 through July 6, 2026 — beating the $70.59 billion VanEck Semiconductor ETF by nearly 37 percentage points over the same window — and it has since given back close to a fifth of its value.

The drawdown is not idiosyncratic. The iShares semiconductor fund entered bear market territory earlier this month after falling more than 20% from a record high set in June. South Korea's Kospi retreated over 4% on July 20 to 6,525 points. The sector has been in a genuine reset since the start of the second half.

What makes PSI worth analysing rather than simply owning is why it outperformed on the way up and how its construction behaves on the way down. The answer to both is the same: it barely owns Nvidia.

The fund tracks the Dynamic Semiconductor Intellidex Index, a quantitative model that ranks US semiconductor companies on momentum, quality, value and management factors rather than market capitalisation. That methodology assigns Nvidia a 4.29% weight. The cap-weighted alternative assigns 18.41%.

Nvidia is up roughly 5% year to date while the broader complex compounded in triple digits. That single spread explains most of the performance gap between the two funds.

Thursday's tape was a live test. STMicroelectronics collapsed 14% to 17% on soft third-quarter guidance. Micron held $964.42, up 0.4%. Nvidia slipped 1.3% to $209.26 and AMD 1.2% to $545.97. SK hynix's US shares gained 2.8%. PSI's mid-cap, equipment-heavy construction outperformed the index it sits inside.

Intel reports after the close, and it is precisely the profile this fund's model is built to capture.

The Methodology: A Rules-Based Active Strategy in a Passive Wrapper

Understanding PSI requires accepting that it is not an index fund in the conventional sense.

The Dynamic Semiconductor Intellidex Index evaluates companies against investment merit criteria including fundamental growth, stock valuation, investment timeliness and risk factors. In practice that means momentum, quality, value and management-quality screens applied to a universe of US semiconductor companies, with weights assigned by rank rather than by size.

The fund invests at least 90% of total assets in index constituents. The index comprises roughly 30 US semiconductor companies engaged principally in the manufacture of semiconductors. The fund is non-diversified by regulatory definition, permitting larger single-name concentrations than a diversified fund could hold.

The output is a portfolio of 31 to 33 holdings with the top ten representing roughly 50.19% of assets — individual positions in the 5% to 7% range rather than the 15% to 20% concentrations market-cap weighting produces.

As of July 1, the top five holdings were Applied Materials at 6.87%, KLA Corporation at 6.48%, Lam Research at 5.80%, Micron Technology at 5.57% and Advanced Micro Devices at 4.98%. Three of those five are semiconductor capital equipment makers, together representing roughly 19% of the fund.

The March composition shows the rebalancing mechanism at work. At that point the top ten read Micron 5.41%, KLA 5.32%, Broadcom 5.10%, Applied Materials 5.05%, AMD 5.04%, Lam Research 5.00%, Nvidia 4.79%, Texas Instruments 4.69%, Tower Semiconductor 3.78% and nLIGHT 3.33%.

Compare the two and the model's behaviour is visible. Applied Materials moved from 5.05% to 6.87% and KLA from 5.32% to 6.48% as their momentum and quality scores improved. Broadcom — which gained just 8% this year — and Texas Instruments dropped out of the disclosed top five.

Sector allocation runs approximately 97.3% technology and 2.7% industrials.

The cost is 0.56% annually against 0.35% for the VanEck fund and roughly 0.34% for the iShares alternative. PSI charges 21 to 22 basis points more for a strategy that outperformed by 3,700 basis points through early July.

That trade has worked spectacularly. The structural question is whether a momentum-ranked model helps or hurts once the sector has already corrected 19%.

The Nvidia Underweight Is the Entire Story

Every semiconductor ETF decision in 2026 reduces to one question: how much Nvidia do you want?

The answer has been counterintuitive. Nvidia — the largest semiconductor business on earth by market value and the company at the centre of the artificial intelligence buildout — has been the laggard of its own sector, up roughly 5% year to date while memory, analog and equipment names compounded at triple-digit rates.

That inversion scrambled the ETF league table entirely. Funds designed to maximise exposure to the AI winner delivered the least. Funds that dilute the megacaps across a broader base delivered the most.

The weight comparison makes the mechanism explicit. The VanEck fund carries 18.41% Nvidia. PSI carries 4.29%. The SPDR equal-weight alternative spreads 53 holdings on a modified equal-weight basis with its top ten at just 28.18% of assets. The iShares product applies an 8% single-stock cap.

Broadcom compounds the problem for cap-weighted funds. A top-five position in both the VanEck product and the PHLX replicant, it gained just 8% this year and dragged alongside Nvidia.

There is a second structural quirk. PSI holds zero shares of Taiwan Semiconductor. TSMC's American depositary receipt status bars it from PSI's US-focused index despite a $2.34 trillion market capitalisation and a 49% gain in 2026. The fund owns chip designers like AMD and MaxLinear rather than the foundry that manufactures for them.

The uncomfortable implication for anyone buying at $152.85: this outperformance is a bet against Nvidia and Broadcom mean-reverting. If the two largest semiconductor companies in the world catch up to a sector that ran 100% without them, PSI's methodology becomes a headwind immediately — and it does so mechanically, because the index reweights on rank rather than holding winners.

The counterargument is credible. Capital equipment and memory are where AI capital expenditure physically lands. Alphabet raised 2026 capex guidance to $195 billion to $205 billion this week. TSMC lifted its own to $60 billion to $64 billion. That money buys deposition tools, etch systems, metrology and memory before it buys a single graphics processor.

PSI is a bet on the picks-and-shovels layer rather than on the end product.

The 19% Drawdown Is the Part Nobody Discusses

The performance headlines describe a fund that doubled. The 52-week range describes a fund in correction.

PSI reached $189.24 at its high and trades at $152.85 today — a decline of 19.2%. It closed at $158.01 on July 2 after falling 9.40% in a single session, lost another 2.67% on July 15, and has drifted lower since.

That path matters more than the endpoint for anyone entering now. An investor who bought at the high is down nearly a fifth. An investor who bought at the $57.36 low is up 166.5%. The fund's own literature explains the volatility: PSI dedicates the majority of assets to medium and small capitalisation companies, meaning it will be more volatile than a traditional large-cap product.

Combine mid-cap exposure with 31 holdings and a 50.19% top-ten concentration and the result trades like a leveraged single-stock bet on one industry.

Thursday's session illustrated it in miniature — a $4.30 range from $151.10 to $155.40 on a fund that closed at $152.85, or 2.85% of intraday travel on a day with no fund-specific news.

The broader complex has been no calmer. The iShares semiconductor fund fell more than 20% from its June record into bear market territory, dropping 3.5% in one Friday session after having been up 89% year to date. The Roundhill memory-focused product dropped over 6% in a single pre-market session.

Liquidity amplifies it. Average daily volume of roughly 461,000 shares on a $3.82 billion fund holding mid-cap semiconductors means both the fund and its investors face execution costs that never appear in the expense ratio.

The practical guidance follows. This is not a core holding for a portfolio that cannot absorb a 20% drawdown inside a month — because that has already happened once this quarter. It is a tactical sector expression and should be sized as one: smaller positions with wider invalidation than an investor would apply to a broad technology fund.

The one thing the drawdown does improve is entry. Buying a momentum-ranked sector fund 19% off its high is a materially different proposition to buying it at the peak.

The Industry Data Behind the Move Is Genuinely Extraordinary

The 2026 performance was not a valuation artefact. The underlying industry numbers are the strongest on record.

Global semiconductor sales reached $298.5 billion in the first quarter of 2026, up 25% from the fourth quarter of 2025. Monthly sales in May reached $120.6 billion — an increase of 104.1% from May 2025 and the highest monthly total the industry association has ever recorded.

A sector doubling year over year on monthly revenue is not a cycle. It is a step change.

The full-year projection reflects it. Global semiconductor industry sales are now expected to reach $1.5 trillion in 2026, hitting that milestone earlier than previously forecast, fuelled by demand for AI infrastructure and accelerated computing platforms.

Invesco's own head of thematic and specialty product strategy framed the driver plainly: artificial intelligence remains the dominant structural driver across the semiconductor value chain. The operative phrase is "value chain" — not the chip designers alone, but the entire stack from equipment through memory through analog through packaging.

Memory illustrates the intensity. Analysts covering the space expect memory contract prices to rise at least 25% from the second quarter to the third, with data-centre-driven shortages showing no sign of abating. The chairman of one of the largest memory producers called current pricing abnormally high and suggested the industry prioritise expanding supply over harvesting the shortage.

Micron traded at $964.42 on Thursday. Sandisk changed hands at $1,570.04. Those are prices that would have been unimaginable eighteen months ago for businesses long treated as commodity cyclicals.

Capital equipment is PSI's largest overweight and the other primary beneficiary. Every gigawatt of new fabrication capacity requires deposition, etch, lithography support and inspection tooling, and order books at the major equipment makers extend years out.

The gap between that industry data and a 19% drawdown in the fund tracking it is the central tension. Either the equities got ahead of even a doubling revenue base, or the market is pricing the end of the capital expenditure cycle that funds it.

Thursday offered evidence for the second reading, and it came from an unexpected direction.

Alphabet's Capex Print Cuts Both Ways for This Fund

The most important semiconductor news of the week was not made by a semiconductor company.

Alphabet raised its 2026 capital expenditure guidance to $195 billion to $205 billion — up $15 billion at the midpoint from a prior $180 billion to $190 billion range and well above the $186 billion to $188 billion analysts had modelled. Second-quarter capex alone hit $44.9 billion, and free cash flow turned negative at minus $5.9 billion.

The stock fell roughly 5%.

For PSI's holdings, that guidance is unambiguously good. Two hundred billion dollars of infrastructure spending becomes revenue for equipment makers, memory suppliers and analog vendors regardless of which architecture ultimately wins. Applied Materials, KLA, Lam Research and Micron — 24.7% of the fund between them as of July 1 — sit directly in that flow.

Asia read it that way. The Kospi rose as much as 3.5% on Thursday and SK hynix's US-listed shares gained 2.8% to $169.88 as investors bet Korean chipmakers capture a share of the spending.

For the sector's multiple, it is considerably less good. The market spent Thursday repricing the artificial intelligence capital expenditure complex on the conclusion that hyperscalers are spending faster than they can convert. Tesla confirmed capex above $25 billion and burned $1.09 billion in free cash flow the same evening. Two of the largest technology companies in the world printed negative free cash flow on the same night.

If investors conclude that spending is unsustainable, they mark down the entire supply chain in anticipation — which is precisely what a 19% drawdown in the best-performing chip fund of the year looks like.

The resolution arrives next week. Microsoft, Meta and Amazon report, and their capital expenditure commentary determines whether the order books PSI's holdings are working through keep filling.

That is the single most important variable for this fund over the coming month, and it has nothing to do with semiconductors.

The Street Is Split and the Disagreement Is Substantive

Two major research houses published diametrically opposed calls on the sector this month.

One large US bank argued the chip sector has had a rough start to the second half but that it is time to buy the dip, framing a summer buying opportunity. Another said the weakness is the beginning of something harder, warning the sector faces a difficult remainder of 2026.

The bearish case carries a specific structural argument: the semiconductor sector is undergoing a reset and has a documented tendency to underperform in the third quarter. Seasonality here is real — inventory builds ahead of the holiday cycle, guidance resets after mid-year, and capital equipment orders lump around fiscal calendars.

The chief US equity strategist at the bearish house has framed the broader dynamic as investors rotating out of the year's biggest winning technology trades, with hyperscalers stabilising while semiconductors face pressure.

That call has been directionally correct so far. PSI is 19.2% off its high. The iShares fund fell into bear market territory. The Kospi dropped 4% in a session.

The bullish case rests on the industry data — record monthly sales, a $1.5 trillion annual run rate arriving ahead of schedule, memory pricing up 25% quarter over quarter — and on the argument that a 19% correction in a sector doubling its revenue is an entry rather than a warning.

Thursday's individual moves supported both readings simultaneously. STMicroelectronics collapsed 14% to 17% on third-quarter guidance of $3.70 billion against a $3.72 billion to $3.79 billion consensus, with EBITDA of $679 million missing a $797.7 million estimate — while raising its data-centre revenue ambition to more than $1 billion in 2026 and above $2 billion in 2027.

A company simultaneously missing near-term guidance and raising its structural AI outlook is the sector in one datapoint. Even after the collapse, that stock remains up more than 110% in 2026.

The disagreement resolves over the next eight weeks, and PSI is the highest-beta way to be wrong about it.

AUM Has Grown Six-Fold and That Changes the Product

The asset base tells a story about how the market found this fund, and it introduces a consideration that did not exist a year ago.

PSI carried approximately $587 million in assets in May 2025 and roughly $727 million in August 2025. By March 2026 it stood near $1.29 billion. By July 1 it had reached approximately $2.86 billion. Market capitalisation now sits at $3.82 billion.

That is a six-fold increase in roughly fourteen months. Price appreciation of 150.93% over twelve months explains a meaningful share, but the remainder is genuine net creations — investors discovering that the second-tier semiconductor ETF was outperforming the flagship by nearly forty percentage points.

Note the sequencing carefully: assets rose from $2.86 billion on July 1 to $3.82 billion now, while the price fell from $158.01 to $152.85. Assets grew as price declined, which means net creations of substantial size have arrived during the drawdown.

That is a genuine institutional-behaviour signal, and it mirrors the pattern visible in crypto ETF flows this month — allocators buying a falling price rather than redeeming into it.

Rapid asset growth in a concentrated mid-cap strategy carries a risk cap-weighted funds do not face. When a rules-based index assigns 6.87% to a company on a momentum rank rather than on market capitalisation, and the fund tracking it grows six-fold, the fund's ownership of that company's float grows correspondingly. At quarterly rebalance, moving a position from 4% to 7% becomes materially larger relative to the underlying stock's daily volume.

That is manageable at $3.82 billion. It becomes a performance drag well before it becomes a capacity problem, through rebalance slippage that quietly erodes the tracking advantage.

The comparison is instructive. The VanEck fund manages $70.59 billion against holdings that are among the most liquid equities in the world. PSI manages $3.82 billion against a portfolio weighted toward mid-caps including names like Tower Semiconductor and nLIGHT.

The fund carries a Morningstar Silver Medalist Rating assigned quantitatively in May 2026, and executed a 3-for-1 share split in July 2023.

What PSI Owns That the Big Funds Do Not

The differentiated exposure is what an investor is actually buying, and it is worth cataloguing precisely.

The capital equipment overweight is the largest structural difference. Applied Materials at 6.87%, KLA at 6.48% and Lam Research at 5.80% together represent roughly 19% of the fund. In a cap-weighted alternative those three collectively carry a considerably smaller weight, because Nvidia, Broadcom and TSMC absorb the top of the portfolio.

Equipment is the purest expression of the capital expenditure cycle. When hyperscalers commit hundreds of billions to infrastructure and TSMC guides capex to $60 billion to $64 billion, the equipment makers book orders regardless of which chip designer wins the architecture race.

Memory is the second overweight. Micron at 5.57% is a substantially larger position than its market capitalisation would justify under cap weighting, and memory has been the strongest sub-sector of 2026 with contract prices expected to rise at least 25% quarter over quarter.

The mid-cap tail is the third differentiator. Tower Semiconductor at 3.78% and nLIGHT at 3.30% as of March are names that do not appear in the largest funds at meaningful weight. Those holdings are the source of both the excess return and the excess volatility.

What PSI lacks is equally important. No Taiwan Semiconductor. Minimal Nvidia at 4.29% against 18.41% in the cap-weighted alternative. Reduced Broadcom exposure after the model cut weight following underperformance.

The absence of Intel from the disclosed top five as of July 1 is worth flagging, because Intel is up 163% in 2026 — the best-performing large-cap chip name of the year — and reports after Thursday's close. Consensus sits near $14.42 billion in revenue and roughly $0.22 in adjusted earnings, with data-centre revenue needing approximately $5.56 billion and options pricing an implied move of about 15%.

A US-domiciled advanced manufacturer with extreme momentum and improving fundamentals is exactly the profile the Intellidex model is designed to capture. Whichever way that print lands, it will show up in the next rebalance.

An investor buying PSI is expressing a specific view: that the upcycle is broad rather than concentrated, that equipment and memory capture more of the economics than the headline designers, and that a quantitative rank-based model keeps rotating toward whatever works.

The Risks Are Concentration, Momentum Reversal, Cost and Duration

Five risks deserve explicit weight at $152.85.

Concentration is the first. Thirty-one holdings, a top ten at 50.19% of assets, weighted toward mid-caps, in a single sector, in a non-diversified fund. There is no meaningful diversification here. A price-to-earnings ratio of 65.96 as of July 1 confirms the fund is not being valued on current earnings.

Momentum reversal is the second and most probable. The Intellidex methodology ranks partly on momentum, which means it systematically increases weight in names that have already run. That amplifies returns in a trending market and produces the worst outcome at an inflection — maximum exposure to previous winners just as leadership rotates. If Nvidia and Broadcom mean-revert upward while equipment and memory correct further, PSI underperforms cap-weighted funds by as much as it has outperformed them.

Cost is the third and it compounds. At 0.56% against 0.34% to 0.35% for the largest alternatives, PSI must outperform by more than 20 basis points annually simply to break even on fees.

Liquidity is the fourth. Roughly 461,000 shares of average daily volume on a $3.82 billion fund holding mid-cap semiconductors means execution costs that never appear in the expense ratio, and they widen in exactly the volatile sessions when investors want to transact.

Duration is the fifth and it is the live macro risk. Thursday had Brent crude at $100.64, the US 10-year yield at 4.695% — the highest since January 2025 — the 30-year above 5%, and September Fed hike odds near 78%. Semiconductors are the longest-duration equity exposure available, and rising real rates compress the multiple applied to earnings arriving years out. A sector at 66 times earnings heading into a possible hiking cycle carries two risks at once.

The offsetting fact remains the industry data. Monthly sales up 104.1% year over year to a record $120.6 billion, tracking toward $1.5 trillion for the year ahead of schedule, is not a sector in trouble.

It may simply be a sector whose equities priced too much of it too quickly — which is what a 19.2% drawdown from $189.24 to $152.85 looks like while the fundamentals keep improving.

How PSI Compares Across the Category

For anyone choosing between the five main semiconductor ETFs, the differences reduce to weighting methodology and nothing else. All five fish from the same pond of roughly 25 to 50 chip names.

The VanEck product is the largest at $70.59 billion, tracking a market-cap-weighted basket of the 25 biggest US-listed names, charging 0.35%, with Nvidia near 18.41%. It returned 66.2% year to date through July 6 and is the default institutional choice.

The iShares alternative applies an 8% single-stock cap to the ICE Semiconductor Index at roughly 0.34%, with holdings distributed more evenly across Micron, AMD, Marvell, Intel and Broadcom. It manages just under $43 billion and was up 89% at one point before falling into bear market territory.

The SPDR equal-weight fund spreads 53 holdings with its top ten at 28.18% — the most diversified structure in the group and the one philosophically closest to PSI without the factor screens.

The Invesco PHLX replicant fully tracks the classic semiconductor benchmark. A VanEck fabless-focused product returned 58.48% year to date through July 6 by excluding manufacturers entirely.

PSI sits between the equal-weight and cap-weighted extremes, using factor ranks rather than size or equal division. It is the only fund in the group whose weights are determined by a quantitative model of company quality.

Performance dispersion across those five reached nearly 40 percentage points through July 6 — the widest spread the category has produced — entirely attributable to Nvidia's underperformance in a year when everything else compounded.

The practical guidance: an investor who believes the AI trade broadens should own PSI or the equal-weight alternative. An investor who believes Nvidia reasserts leadership should own the cap-weighted product.

For a pure-sector allocation with no view on internal leadership, the equal-weight structure at a lower fee is arguably the cleaner instrument than PSI's factor overlay, which introduces active risk without active management accountability.

At a 19% drawdown, the argument for PSI over the alternatives is that its holdings — equipment and memory — are the parts of the chain with confirmed order books, while the cap-weighted funds remain hostage to two megacaps that have not participated all year.

Levels: $155.40 Is the Session Ceiling, $151.10 Is the Floor

The technical structure is defined by three reference points from the live tape and two from the annual range.

Immediate resistance is $155.40, Thursday's session high, followed by Wednesday's $153.47 close which the fund is currently trading beneath. Above those, the $158.01 close from July 2 marks the level PSI held before the current leg lower, and the 52-week high of $189.24 sits 23.8% above spot.

Immediate support is $151.10, Thursday's session low, backed by the $151.23 opening print. Those two levels sit within thirteen cents of each other, which makes the $151.10 to $151.23 band the line that decides whether this session's recovery from the lows holds.

Below that, there are no recent reference points until the fund's ascent through the spring, and the 52-week low of $57.36 is a level from a different market regime entirely.

The intraday range of $151.10 to $155.40 — 2.85% — is the number to internalise for position sizing. On a $3.82 billion fund with 461,000 shares of average daily volume, that is normal rather than exceptional, and stops placed within 1.5% of entry will not survive an ordinary session.

The relative levels matter as much as the absolute ones. PSI at 19.2% off its high sits in correction territory but has not yet met the 20% bear-market threshold that the iShares alternative crossed earlier this month. That is a distinction without much practical difference, but it does mean this fund has held up marginally better than the cap-weighted competition through the drawdown — which is the methodology working in both directions.

For anyone measuring the entry against the sector cycle, the useful frame is that the fund is 166.5% above its 52-week low and 19.2% below its high. Whichever of those numbers feels more relevant is effectively the investment view.

The 2.85% daily range is the price of holding it either way.

Forecast: Range Between $145 and $165 Until the Hyperscalers Report

The base case into month-end is continued volatile consolidation between roughly $145 and $165, with the bias determined by next week's capital expenditure guidance rather than by anything semiconductor-specific.

The evidence: record industry revenue with monthly sales up 104.1% year over year, memory pricing rising 25% quarter over quarter, and net creations arriving during a drawdown — set against a 19.2% correction from the high, Brent at $100.64, a 10-year yield at 4.695%, and a market actively repricing the AI capital expenditure complex.

The bullish scenario requires reclaiming $155.40 and then $158.01 on volume. Triggers: a clean Intel print after Thursday's close validating US advanced manufacturing, Microsoft, Meta and Amazon confirming rather than moderating capital expenditure next week, or a Middle East de-escalation that pulls yields and September hike odds lower. That path opens $165 and, on a genuine sector re-rating, the $189.24 high — 23.8% above spot.

The bearish scenario activates on a close below $151.10. Triggers: a weak Intel guide, any hyperscaler signalling capital budget review, a hawkish FOMC statement on July 28-29, or confirmation that the documented third-quarter seasonal underperformance is materialising. Below $151, the fund has thin recent structure and $140 becomes the realistic objective, which would take the drawdown from the high to 26%.

The calendar is dense and front-loaded. Intel reports after Thursday's close with a 15% implied move. The FOMC decides July 28-29. Microsoft, Meta and Amazon report across the following week. The quarterly index rebalance will reshuffle weights based on momentum and quality scores calculated through a correction.

What would change the framework entirely is Nvidia. At roughly 5% year to date against a sector that compounded in triple digits, the largest chip company in the world is either broken or coiled. If it catches up, PSI's 4.29% weight becomes a liability immediately and the performance gap closes from the wrong direction.

PSI's methodology captured this upcycle better than any alternative in the category. The test now is whether it navigates the first real correction without surrendering the outperformance — and at $152.85, that test is roughly halfway through.

That's TradingNEWS