Markets

We still think AI is a buy. After the midterms the semiconductor index fell 11.4% while the S&P 500 lost 2.5%, in our view an overreaction that fell on the chip names. October payrolls rose 165,000, oil stays high with the Iran war unresolved, Alibaba and Tencent gained, NVIDIA trades at 15.4 times forward earnings, and both SpaceX and Tesla have special committees studying a combination. KXCO's AI ontology shows the relationships behind the NVIDIA, China and SpaceX calls.

After the midterm sweep, Democrats want to slow America's AI build-out. The Sanders and Ocasio-Cortez bill would freeze building and upgrading of AI data centres nationwide, and a superintelligence ban Rep. Yassamin Ansari wants atop the agenda would pause every model trained with 10^25 operations or more. Governors-elect in Texas and Georgia back pauses, and Goldman Sachs puts 45% of planned capacity in Democrat-run states backing moratoriums. Shayne Heffernan: AI computing follows the power.

KXCO's FDA ontology ranked INO-3107, from Inovio (INO), first of the FDA decisions due in the next 90 days, from 21 October 2026 to the decision. The FDA granted accelerated approval on 30 October; the news reached the market after the close. From its $0.60 close that day, Inovio traded as high as $2.71 after hours, up 351.7%, and closed 2 November at $1.89, up 215.0%. Both published research books closed ahead of XBI. On our radar: BBP-418 (BBIO), Bezuclastinib (COGT) and Relutrigine (PRAX).

Shayne Heffernan rates seven AI stocks buys for the week of 1 November 2026: Nvidia, Broadcom, TSMC, Microsoft, Oracle, Palantir and ASML. The latest results put the contracted backlog of the four largest cloud sellers at about $2.48 trillion, and Amazon, Alphabet, Microsoft and Meta plan about $720 billion to $740 billion of capital spending in 2026. Every call links to KXCO's AI ontology, updated on 1 November with the quarter's results, power and grid data and new defence contracts.

From 21 to 30 October 2026, the first screen of KXCO's FDA ontology at kxco.ai/fda ranked Inovio's INO-3107 #1 in the next 90 days by potential return, at +112% on the first day and +103% from 22 October on its own sourced inputs. On 30 October the FDA granted it accelerated approval, on its target date. The room's watch flagged the approval that night, and the record took in the FDA's own letter 19 minutes later.

Markets debate whether artificial intelligence (AI) is overvalued. Shayne Heffernan argues the more material risk is underestimation. Alphabet's capital expenditure reached $51.4 billion in the third quarter, Google Cloud revenue rose 103% to $30.7 billion and the Cloud backlog stands at $561 billion, per its 28 October results and earnings call. Jurisdictions that block data centres do not remove the demand: they import the capability later, at a higher price.

Shayne Heffernan assigns a 70% probability to the authorisation of a significant US strike package against Iran within weeks. Two carriers and an amphibious group are in the Arabian Sea with more reported due by the end of November, Axios reported that US Central Command (CENTCOM) was told to prepare for a possible resumption of major combat, and officials told CNN forces are prepared to act within 96 hours of an order. Here is what a strike means for oil and defence stocks.

The AI market is now decided by physical capacity, and this year it went to war. US forces used Palantir's Maven against 13,000 targets in the 38 days of Operation Epic Fury, per the Pentagon's AI chief, while hyperscalers called for about $730 billion of 2026 capital spending, per CNBC. Shayne Heffernan reads demand, chips, power, the Gulf escalation around the 3 November midterms, and the defence AI companies: Palantir, SpaceX and xAI, Anthropic, OpenAI, Alphabet, Anduril and Shield AI.

The United States still leads in frontier model capability, total AI compute and quantum error correction, while China has narrowed the gaps in model performance, open-weight adoption and domestic chip volumes, and holds a structural advantage in power for data centres. Shayne Heffernan sets out the developments from July to October 2026 across models, compute, quantum and electricity, and how fast China is catching up.

The constraint on AI is no longer ideas or data but compute and the power to run it. Frontier training compute has grown about five times a year since 2020, and the leading AI supercomputer of June 2030 is on course for about 2 million chips and 9 gigawatts. Shayne Heffernan sets out the 2030 arithmetic, the power problem, who has already contracted the gigawatts, and how GPUs, custom ASICs, wafer-scale engines and quantum machines divide the work.

On 19 July 2026 the Round Table's AI room published its first Analyst Outlook. To the 16 October 2026 close its 13 names rose 2.6% on an equal weight against 2.1% for QQQ, per StockAnalysis closes, and the six it opened with 35% or more of upside rose 8.6%. This piece lists the six names up more than 5%, the ten names I rate as opportunities now, the month's results in AI and quantum computing, and the official data showing that four in five firms have not yet started using AI.

SpaceX did not add an AI business. A company whose processes run on compute built its own computing plant, then built the power to run it. The 10-Q reports 1.4 gigawatts of nameplate compute at 30 June 2026, four leases worth about $41 billion a year and $100 billion of cash. The target is closer to 10 gigawatts than 5 by the end of 2027. This piece sets out the build sequence, the four layers of power in Memphis, and why Terafab is the act after 2027, not the road to it.

There is no AI bubble. A bubble is capacity nobody uses, demand nobody signs for and a glut when the building stops. KXCO's AI sector ontology, refreshed on 7 October 2026, finds the reverse: adoption rising on six independent measures, about $2.35tn of contracted backlog at the four largest sellers of compute per their own reports, and supply rationed at every layer. It keeps the bear case in the same record, 20 circular capital claims, and answers it.

The KXCO Biotech Ontology's board carries 11 pending FDA filings, due by 23 April 2027. Ranked on approval standing, the top five are BridgeBio's BBP-418, Cogent's bezuclastinib, Belite Bio's tinlarebant, Novo Nordisk's CagriSema and BioMarin's vosoritide. Ranked on consensus upside, they are Cullinan, Savara, Intellia, Mineralys and Pharvaris. No name is on both lists. This report covers BridgeBio, Cogent, Mineralys, Pharvaris and Savara.

Pete Hegseth's Autonomous Warfare Command is a four-star proposal that needs Congress by 1 October 2027. Until then Project Agincourt holds the work, and the money moves through Drone Dominance, the Army's counter-drone contracts and DIU. The software layer, Palantir's Maven, Anduril's Lattice and Shield AI's Hivemind, is the one the KXCO AI Sector ontology had mapped before the speech. After Minab, its hardest problem is the record, not the model.

In June, three crypto platforms cancelled SpaceX pre-IPO tokens because they could not get the shares. In 2026 the SEC wrote down why: a tokenised share must carry the same dividends, votes and liquidation rights as the share, traded only by permissioned participants. Every one of those is a question for the share register. The Round Table maps what the rules ask, where KXCO keeps each answer, and the listing routes to tokenised-asset exchanges.

Your agent already acts in a legal system that was built for people. No new statute is needed for a payment it releases to be an act of your institution. Twelve months later a court asks who acted, by what right, and what can be reconstructed, and an export from a product you do not own is not a file. KXCO writes the act quantum-signed, time-stamped and marked on chain before the dispute. Knightsbridge Law sits on every Round Table ontology, so counsel is at the table before the act.

The people who will be blamed for what AI does are not in the room where its safety is argued. A jailbreak escapes a policy. The failure that costs money is an action that escapes an owner, a payment sent or a record altered. Tokens are not authority. Intelligence got cheap and judgment did not, and there is no organ of standing you can copy into a model. Shayne Heffernan on the objection worth taking seriously, and the questions to take into a vendor meeting.

Elon Musk is sleeping in an Airstream beside a warehouse in Whitehaven with MACROHARD painted on the roof. The machine next to it is a two-state compute campus carrying about 1.29 gigawatts of IT load, roughly 670,000 accelerators and 48.7 billion dollars of modeled capital, per Epoch AI cards of 16 September 2026. A field brief on the buildings, the silicon, the turbines, the batteries, the tenants who now hold the hours, and what the compute is actually worth.

Quantum is accelerating. Not toward a machine that breaks RSA next quarter, which is still five orders of magnitude away on the first honest cross-platform yardstick the field has ever had, but toward foundries, clouds, logical qubits and government deadlines that are already fixed. Two clocks are running. Only one of them is slow, and it is not the one that decides what a bank, a court or a ministry has to do this year.

Commentators pointed at a flat Nasdaq, a few missed prints, the cost of a megawatt and a safety letter, and called it a slowdown. It is not. Between 12 August and 16 September 2026 nine labs shipped frontier or near-frontier models, and the contest moved from one leaderboard to two stacks. What actually slowed is accountability. Enterprises still cannot put any of it into a court file. That gap, not the capability curve, stops deployment inside banks, hospitals, courts and ministries.

People treat the Jack Sparrow line as a joke. The attribution is a joke. The sentence is not. Forty years in markets says the same thing Epictetus said in 125 CE and Robert Merton named in 1948: the event is finite, and the story you appoint to govern it is not. This essay walks the quote back to its actual sources, draws the loop that turns a feeling into an order, and sets out the four places a rule written in advance cuts the loop. Faith is not a hedge. It is a stance.

China is no longer trying only to train a model that looks like an American one. It is trying to own the stack that turns tokens into factories: open-weight models cheap enough to run at home, accelerators that do not need an Nvidia licence, robots that walk into a BYD plant, quantum machines that list on the STAR Market, and a power system that can add the generating capacity of a mid-sized country in a single year. The models, the tickers, and the 27 claims that cross the border.

The first week of September 2026 compressed a year of argument into seven days. GPT-6 Astra landed at the Critical cybersecurity level of OpenAI's own Preparedness Framework, Nvidia agreed to buy Hugging Face for $12.93 billion, Washington took minority equity in three quantum companies, and Mistral closed the largest equity raise in European technology history. A field report on the breakthroughs, the tickers, and a map that has split three ways.

Eight molecules carry an FDA label. Six carry a nonbinding committee vote. The gray market sells the second as though it were the first. A field report on the longevity pipeline, the capital behind it, and the listed names that matter.

Twenty FDA approval actions so far in 2026, eight of them inside a sixteen-day window in August. A first-in-class RAS inhibitor that doubled survival in pancreatic cancer, an oral GLP-1 cleared in fifty days, a first gene therapy for GSDIa, and seven AI platforms with nine assets in humans and nothing yet approved.

Nvidia is the most connected name on the KXCO Ontology and is not a chokepoint. Cadence carries two claims and is. That distinction decides the whole ownership question in semiconductors right now.

SpaceX listed in June, absorbed xAI in February and now rents compute to the labs it competes with. Elon Musk holds about 82 percent of the vote there and only the CEO seat at Tesla, which settles the direction of any merger. Three Neo4j figures test the argument, and on one point the graph disagrees.

A Sunday strike on Larak Island put the Hormuz premium back on, Warsh put a September hike back on the table, and Friday NFP decides both. Full daily-chart levels for stocks, gold, silver, oil and Bitcoin, plus three figures from the KXCO ontology showing what actually breaks the AI trade.

Five weeks before Nvidia reported a $96.2 billion quarter and added $442 billion in a day, the KXCO ontology had it at $203, Strong Buy, for a reason anyone could read. Here is the full 13-name scorecard from 24 July to the 27 August close, the Cypher queries behind the call, and the misses.

A dated, checkable one-month scoreboard on the KXCO showcase ontology, the Nvidia move up the stack, Alibaba's $10.2bn raise, the Trump SpaceX disclosure, and the seventeen AI majors ranked by implied upside at the 24 August close.

Core PCE and Q2 GDP land at 08:30 on Wednesday, Nvidia reports that evening, and Warsh gives his first Jackson Hole keynote on Friday. Verified levels, the 16-name AI buy list, and three graph analyses of where the real chokepoints sit.

Memory supply is growing about 20% a year while AI demand climbs 200%. Shayne Heffernan on the agentic workload behind the shortage, Micron's real fab timelines, the five listed names carrying the trade, and why quantum computing increases demand for classical memory rather than replacing it.

Musk's August 14 statement that orbital compute will be the only way to scale AI by 2029 caps a year in which SpaceX moved space-based data centers from concept to regulatory filings, a named Nvidia partnership and factory construction.

ARK fully exited Twilio in 2024, so the old $44.7 million story is history. The capital is now flowing into SpaceX, wafer-scale AI silicon, GPU cloud and stablecoin rails. Here is what Cathie Wood is actually buying as of mid-August 2026, what she is trimming, and whether the math still holds.

Central banks measure power in basis points. The Intelligence Age measures it in terawatt-hours. A country-by-country audit of who generates the electrons, who is building the next grid, what industrial power actually costs, and why China added a Germany-sized grid in a single year.

Cerebras core revenue doubled and its fast inference cloud nearly quadrupled in Q2 2026, with more than 600 MW of data-centre capacity under contract. A deep dive into the wafer-scale silicon, the software stack, the 25 kW thermal problem, and the prefill-on-GPU, decode-on-wafer split now entering production.

Elon Musk said energy would become the de facto currency. He was describing a physical constraint, not a metaphor. Firm power delivered to the rack is now the limiting reagent in the AI build-out, and roughly 158 GW of data-centre capacity still has to be energised by 2030.

The AI arms race has moved from what the technology can do to who controls the compute, the data and the distribution. Anthropic now has five direct competitors on one cap table at a $965bn post-money. Meta is commoditising the model layer it does not sell. Nvidia carries 52 dependencies in the KXCO ontology and the sector resolves to nine critical nodes. Here is the state of the AI economy, checked line by line against our live map, and the five value gaps that map is pointing at.

The S&P 500 closed at a record on Friday and the tape is bullish, so the job this week is choosing, not chasing. Palantir is up 28.6 per cent from the KXCO ontology's recorded 15 July snapshot and SpaceX is up 15.6 per cent since we published on 22 July. Here is the full economic calendar for 10 to 14 August, what each print means for positioning, and the fourteen value gaps the ontology's Analyst Outlook layer is pointing at, from Oracle at 70 per cent down to Arm at 2 per cent.