# KXCO's Top 5 AI Sector Picks Right Now, and Where the Value Still Sits

The KXCO ontology recorded a dated snapshot of thirteen AI majors on 15 July. Eleven of thirteen are up, none has been down-rated, and eleven had their target raised. Shayne Heffernan reads what moved, then sets out where the remaining upside actually is.

Canonical HTML: https://www.livetradingnews.com/kxco-s-top-5-ai-sector-picks-right-now-and-the-three-week-scorecard-on-the-last-
Last modified: 2026-08-05

---

By Shayne Heffernan · 2026-08-05
Tags: AI stocks, $ASML, $NVDA, $ORCL, $PLTR, $GOOGL, $INTC, $META, KXCO ontology, semiconductors, EUV lithography, chokepoints, circular capital, hyperscaler capex, AI infrastructure, Shayne Heffernan, John Heffernan, market structure, Q2 2026 earnings, knowledge graph
Signed: ML-DSA-65, anchored on Armature L1.
Nothing in this article is investment advice.

Three weeks ago the KXCO ontology recorded a snapshot of the AI sector: thirteen listed majors, their prices, their valuations and the consensus twelve-month targets attached to them. That snapshot was dated 15 July 2026 and it has not been edited since, because the map keeps its own history rather than overwriting it.

The market has since had one of the more eventful three weeks of the cycle. Q2 earnings landed between 22 and 30 July. It is a reasonable moment to open the old file and see what the structure was pointing at.

## What went up

Every figure below is the recorded 15 July price against the 4 August close. Nothing was selected after the fact.

- **Intel** 73 to 101, up 38.4 per cent
- **Microsoft** 392 to 493, up 25.8 per cent
- **Palantir** 132 to 163, up 23.5 per cent
- **Amazon** 233 to 277, up 18.9 per cent
- **Oracle** 124 to 146, up 17.7 per cent
- **Alibaba** 110 to 129, up 17.3 per cent
- **Apple** 270 to 309, up 14.4 per cent
- **Alphabet** 334 to 378, up 13.2 per cent
- **Baidu** 100 to 113, up 13.0 per cent
- **Arm** 267 to 281, up 5.2 per cent
- **Nvidia** 203 to 212, up 4.4 per cent

Eleven of thirteen rose. The unweighted cohort average is up 13.1 per cent in three weeks.

Palantir is the one worth dwelling on. The July file carried it at $132, and the map placed it at the centre of the government and defence integration layer, connected to the US Department of Defense, ICE, the UK government and the IDF. On 3 August the company reported second-quarter revenue up 93 per cent, with US commercial revenue up 149 per cent, and the stock closed the following session at $163. The structural position the map described turned into the numbers the market repriced.

Intel is the second. The map holds it for the 18A foundry programme, which matters because the sector's single greatest concentration risk is that leading-edge fabrication sits with one company on one island. A credible second source is worth a great deal to everyone in the graph, and the market has begun to pay for the possibility.

## There is no finish line

A twelve-month price target marked after three weeks is not a result. It is a position with eleven months left to run, and the more useful question is not what moved but where the room still is.

On that measure the file has held up better than the price moves alone suggest. Since 15 July:

- **Not one of the thirteen has been down-rated.** Every name still carries a Buy, Moderate Buy, Strong Buy or Hold, exactly as it did in July.
- **Every one still carries positive remaining upside** to its consensus target.
- **Eleven of thirteen had their target raised**, not cut. Oracle went from $198 to $248. Microsoft from $530 to $563. Intel from $89 to $115. Baidu from $143 to $172. Tencent from HK$650 to HK$692.

A stock that has not moved yet has not failed. In several cases it has become a better proposition than it was, because the target rose while the price did not.

## Where the value remains

Four names now carry more remaining upside than they did in July. Three of them are up on the period and one is not, which is rather the point.

**Oracle.** Up 17.7 per cent and yet the remaining upside widened from 60 to 70 per cent, because the target moved from $198 to $248. At $146 it trades on 25 times trailing and 18.1 times forward. Oracle Cloud Infrastructure revenue is up 47 per cent against a backlog around $75 billion. It is the compute provider without a frontier model of its own to defend, which is precisely why the labs will rent from it. The widest remaining gap in the cohort.

**Tencent.** Down 8.3 per cent over the period while its target was raised from HK$650 to HK$692, which took the remaining upside from 20 to 41 per cent. It trades on 16.9 times earnings with a free cash flow yield that stands well clear of its domestic peers, and AI capital spending has doubled. The position did not weaken. The entry improved.

**Baidu.** Up 13.0 per cent with the remaining upside widening from 43 to 53 per cent on a target lifted from $143 to $172. AI now accounts for 52 per cent of total revenue, and the company has its own Kunlun accelerator, which gives it partial insulation from the silicon question that constrains every other Chinese lab.

**Meta.** Down 13.7 per cent on a Q2 earnings miss and a 91 per cent fall in free cash flow, and its remaining upside expanded from 21 to 29 per cent. It holds a Strong Buy with a $757 target against $588, on 22 times trailing and 18.4 times forward, the second cheapest multiple in the cohort. The cash flow fell because capital spending rose to a guided $130 to $145 billion. That is the cost of the position, not the failure of it.

The mirror image is worth stating plainly. The names that ran hardest now have the least room left: Arm at 2 per cent of remaining upside, Apple at 4, Alphabet at 13, Intel and Microsoft at 14, Palantir at 16. Nothing there has broken. The value has simply rotated toward what has not moved yet.

## The five the structure supports now

Ordered by position in the stack, from the bottom up. Prices are the 4 August 2026 close, and the upside quoted is what remains.

### 1. ASML, the root

The single most important finding on the map is not about a model company. It is that the entire sector resolves to one Dutch firm in Veldhoven that makes the extreme ultraviolet lithography every leading-edge AI chip requires. ASML was not in the seed set when the graph was built. The engine found it by following dependencies outward from the companies everyone talks about.

To be precise about what that is: a structural call about concentration, not a claim about the share price. ASML sits at $1,712 with a $651 billion market capitalisation, on 53.7 times trailing and 30.2 times forward, with a Strong Buy and a $2,126 consensus target, some 24 per cent above the current price. It pulled back from its late-June high on concern about Chinese competition in lithography.

Every dollar of the roughly $725 billion the four largest hyperscalers will spend on AI infrastructure this year has to pass through silicon that only ASML's machines can print. There is no second source and no quick way to build one.

### 2. Nvidia, 43 per cent still on the table

Nvidia rose 4.4 per cent over the period and remains the widest gap among the large caps, with a $303 target against $212 and 43 per cent of remaining upside. Trailing revenue is up 70.7 per cent and net income up 107.9 per cent, on 32.5 times earnings and 21.2 times forward.

The map flags it as the sector's single largest point of failure, with 52 entities depending on it. That is the risk and the position in the same sentence. What has changed is the shape rather than the size: AMD's up to two gigawatt arrangement with Anthropic is the first credible second source at the frontier tier, and in late July Nvidia and OpenAI were reported to be discussing a backstop of up to $250 billion allowing OpenAI to raise debt against Nvidia's credit for a ten gigawatt Ohio campus. Every earlier circular deal here was equity. Those talks are reported, not signed, and the map records them that way.

### 3. Oracle, 70 per cent still on the table

Covered above. The neutral landlord, the widest remaining gap, and the clearest example of a position that improved while the price rose.

### 4. Palantir, the integration layer

Palantir is where AI stops being a demonstration and becomes procurement. The map connects it to the US Department of Defense, US Immigration and Customs Enforcement, the UK government and the IDF, and it is one of eight firms holding classified network accreditation. Second-quarter revenue rose 93 per cent, US commercial revenue rose 149 per cent to $764 million, and full-year guidance rose to roughly $8.15 billion.

At $163 it trades on 139 times trailing and 86 times forward against a $189 target, which leaves 16 per cent. That is the thinnest cushion of the five and it is the consequence of the move already delivered. The position is a judgement that government AI budgets are stickier than commercial ones.

### 5. Alphabet, the only full stack

Alphabet owns every layer: its own accelerator in the TPU, its own frontier model, and its own cloud to sell both through. Everyone else rents at least one of those.

Second-quarter revenue was $119.8 billion, up 24 per cent, with Google Cloud up 82 per cent to $24.8 billion and cloud operating income up from $2.8 billion to $8.8 billion. At $378 it is on 19 times trailing earnings, the lowest multiple of the five, with a $428 target. Guided capital spending of $195 to $205 billion is why the multiple sits where it does.

## What would change the picture

The last week of July repriced the chip complex. Nvidia lost $238 billion of market value, SK Hynix $176 billion, Samsung $173 billion and Micron $113 billion, with AMD and TSMC each shedding more than $100 billion. The stated cause was not collapsing demand but a repricing of expectations, on the concern that AI infrastructure spending may be peaking faster than expected.

That is the risk to all of these at once. The concentration that makes the chokepoints valuable on the way up makes them correlated on the way down, and the memory tier fell hardest. Anyone holding this stack is holding one trade, not five.

The second consideration is funding. Combined hyperscaler capital spending of about $725 billion this year, up 77 per cent, is increasingly financed through arrangements where the vendor stands behind the customer. Fifteen such loops now run through this sector, and they are contagious on an unwind.

## Check the working

Every claim above comes out of a map you can open and interrogate. It is free, it is public, and every relationship on it carries a source link, a date and an honest confidence level, so you can walk any number back to the filing or report it came from and disagree with us specifically rather than generally.

Every claim also carries the dates it was true from, which is why a three-week-old snapshot still exists to be checked rather than having been quietly overwritten.

The map is at [kxco.ai/ontology-live](https://kxco.ai/ontology-live/). There is a guide to reading it at [kxco.ai/developers/blog/ontology-live-guide](https://kxco.ai/developers/blog/ontology-live-guide), and background on the method at [kxco.ai/ontology](https://kxco.ai/ontology).

Disclosure: LiveTradingNews and KXCO are part of the same group. The ontology is built by Shayne Heffernan and John Heffernan. The valuation figures quoted are sell-side consensus as at the 4 August 2026 close, collected and dated, not our own forecasts. None of this is investment advice.

---

This Markdown mirrors https://www.livetradingnews.com/kxco-s-top-5-ai-sector-picks-right-now-and-the-three-week-scorecard-on-the-last-. The HTML page is canonical.
Site index for AI clients: https://www.livetradingnews.com/llms.txt