Invesco AI and Next Gen Software ETF
Invesco
Tracks the STOXX World AC NexGen Software Development Index — a global, non-diversified AI/software basket. Despite the 'software' name it is heavily semiconductor/hardware-weighted in practice (~100 holdings, top 10 ~59%).
Top holdings
Memory/HBM
AI accelerators and CPUs
Hyperscaler/AI infrastructure
Google/DeepMind, AI plus cloud
Dominant AI GPU
Korean memory/HBM
Recent moves
Formerly the Invesco Dynamic Software ETF (ticker PSJ); repositioned to IGPT on the STOXX NexGen Software index after close on Aug 25, 2023.
Our take
A blended chip-and-platform basket — its top names (MU, AMD, META, GOOGL, NVDA) span both the AI-hardware and hyperscaler trades, which is why it tracked closer to the semis than to pure-software peers like IGV in 2026. Smallest AUM here; the 2023 PSJ→IGPT rename means its longer track record describes a different strategy.
The name is the trap. A buyer reaching for 'next-gen software' gets a fund whose largest sleeve is memory and logic chips plus two ad-platform mega-caps — the index definition of 'software development' sweeps in the entire hardware value chain.
That mislabel is exactly why IGPT worked in 2026: it sidestepped the pure-SaaS carnage (IGV -12% YTD) because it barely owns pure SaaS. The teaching point marketing won't state: IGPT is not a diversifier against a chip position, it IS a chip position wearing a software label, so pairing it with SMH/SOXX/AIQ stacks the same memory-and-logic bet rather than spreading it. Its 3yr/5yr numbers belong to the old PSJ fund and should not inform a decision on today's strategy.
- 2026-07Update, Jul 2026: the symmetry arrived — barely owning SaaS cut both ways, with the book down ~16% in a month when the S&P 500 was -0.1% and NVDA (+0.3%) and AVGO (+3.1%) held.
Thesis
IGPT is marketed as an AI-and-software fund but its book is a global semiconductor-and-mega-cap-platform basket: 8 of the top 10 are chip/hardware names (AMD, NVDA, MU, SK hynix, INTC, SanDisk, Seagate, MediaTek), with only META and Alphabet as the 'software.' In 2026 it behaved like a semis fund, not a software fund.
Passively tracks the STOXX World AC NexGen Software Development Index (~113 holdings, top-10 ~60%). Formerly PSJ (Invesco Dynamic Software), re-indexed to IGPT after 25 Aug 2023, so its long track record describes a different, US-only quant software strategy. Unlevered — no daily reset or decay.
Assessment
- Chip/mega-cap tilt let it dodge 2026's software de-rating — +50.9% YTD vs IGV -12%, a ~63pt divergence
- Unlevered and broadly held (113 names, top-10 ~60%) — no daily-reset volatility decay of its leveraged AI-cohort peers
- One-ticker blend of AI hardware (AMD/NVDA/MU) and platform demand (META/GOOGL) spanning both sides of the AI trade
- Genuine non-US chip exposure (SK hynix, MediaTek) most US 'AI software' funds lack
- Index reconstitution can shift the semis-vs-software mix with no notice, changing what the ticker actually is
- Smallest AUM (~$1.23B) of the AI cohort here — thinner liquidity than SMH/SOXX/AIQ
- 0.56% fee for exposure largely replicable by cheaper semis ETFs
- 2026-07Update, 31 Jul 2026: the dodge reversed — IGPT fell ~16% in July against the S&P 500's -0.1%, leaving YTD at +45.8%. Set against IGV at -11% to -14% YTD, the gap is nearer 58-60pt than 63.
- 2026-07Update, Jul 2026: the two-sided framing holds by label, not by risk — the hardware sleeve alone drove ~-16% in a month when the accelerator names and the mega-cap platforms were not the problem.
Record
Verified 1yr +78.4% and YTD ~+50.9% (mid-Jul 2026) are a semiconductor-and-memory record, not a software one. Attribution: the memory names (MU, SK hynix, SanDisk, Seagate) rode 2026's HBM/DRAM leg, AMD/NVDA carried the logic side, and META/GOOGL added mega-cap ballast — while near-zero weight in the high-multiple SaaS names that sank IGV (-12% YTD) and WCLD is what it avoided, not what it won on. The gap vs IGV is the entire story: same 'software' label, opposite portfolios, 63-point outcome spread. Pre-2023 track record is the discontinued PSJ strategy and is not comparable.
- 2026-07Update, 31 Jul 2026: the levels have moved — 1yr +70.0% and YTD +45.8%, against +111.5% 1yr at 30 Jun. The memory sleeve reversed in July: SanDisk -46.6%, Micron -28.7%, SK hynix sharply lower.
Risks & fit
- Behaves like a semis fund, so it offers little diversification against an existing SMH/SOXX/AIQ chip holding — correlated, not complementary
- Heavy memory concentration (MU/SK hynix/SanDisk/Seagate) ties it to the cyclical cost-per-bit boom-bust that can invert violently
- Rename discontinuity: screening on its multi-year record measures a fund that no longer exists
- Asia chip exposure (SK hynix, MediaTek) layers Korea/Taiwan macro and FX beta
The 'it's really a semis fund' read breaks if, in a memory-cycle downturn or a software re-rating, IGPT decouples from SMH/SOXX and tracks IGV instead — or if a future index reconstitution swings the book back toward pure software so its correlation to the chip names collapses. Either would mean the label finally matches the holdings.
Someone who wants a single-ticker blend of AI hardware plus mega-cap platforms and understands they are buying semiconductor-and-memory cyclicality, not software — and who is NOT using it to diversify an existing chip position. A poor fit for anyone seeking the pure-SaaS software exposure the name implies.
0.56% expense ratio — above the semis ETFs it most resembles (SMH/SOXX ~0.35%) and above IGV (~0.41%), below the active ARK AI funds (0.75-0.88%). Mid-pack fee for exposure a cheaper semis fund largely replicates.