YvesSaintPige's September 2026 Portfolio Review
September 2026 Portfolio Review
September continued with a solid march upwards of another 10%+ for my holdings. I didn't make any buys or sells during the month, only my regular weekly contributions. $SNOW had earnings at the beginning of the month, $MU had their earnings at the end of the month.
Earlier in the month I did an exercise trying to steelman the bear cases for AI that most concern to me. It is good context for my review of each company and the macro. Link to that post here: https://www.reddit.com/r/GrowthStockInvesting/s/RcjhEQkeTK
Nominal is pure % change, Real takes into account contributions made.
Monthly Gain: Up 13.0% Nominal // Up 12.2% Real
Yearly Gain: Up 75.2% Nominal // Up 50.3% Real
All Time Gain (from 2019): Up 78.5%
$NBIS // Up 20.3% // 45.8% Allocation
Nebius continues to headline the portfolio by a large margin. Fastest growth one can find, in an industry/sector expanding exponentially, with a leadership team who is executing flawlessly. They are a vertically integrated AI cloud company that designs, builds, and operates its own data centers, GPU clusters, and full software stack, renting AI compute to everyone from startups to the biggest companies in the world, Microsoft and Meta.
There was a decent amount of news this month. Starting on 9/8, with Palantir naming Nebius their preferred sovereign AI infrastructure partner, so Palantir's commercial customers can run their own models with control over compute, data, and the models themselves, enabled by Nebius. While there is no dollar value, contract, or concrete timeline, this announcement is further validation that Nebius is trusted by major institutions.
Then, in mid-September, came the huge announcement from major AI executives calling for slower pace in AI development. Dario Amodei initiated the call with a detailed essay, "We Must Pace the Frontier." OpenAI, Google, Microsoft, and other leaders followed in agreement. Many are skeptical that these major AI names are attempting "regulatory capture," trying to close the door behind them to lock in their positions as top dog in the sector, as they face more pressure from potentially open source models. Other people are more inclined to take them at their word and share in the fear that we could irreversibly mess up this moment if we don't take care during AI's initial development. I fall more in the latter's camp, but see both arguments. However, when it comes to Nebius and my other AI stocks, I do not see being cautious and intentional about AI development in any shade of negative light. In fact, I think this intentionality would be much better overall compared to a no-holds-barred race to develop "the best" model at all costs, or reach AGI/RSI/SGI first etc. Bottom line is: this technology is inevitable, so we should move forward with purpose.
That inevitability gained another proof point on 9/23 as Nebius announced another round of Token Factory price increases equaling up 18.3% on average. Nebius also launched a pilot to try a new idea that involves auctioning their endpoints off to let the market set the price for capacity. During the pilot they cleared their highest Blackwell price to date. Demand is sky high. Live compute remains the bottleneck and customers are willing to pay. Also on 9/23, Nebius joined CoreWeave in the Platinum tier of SemiAnalysis' ClusterMAX ratings, a high technical benchmark confirmation. Finally on 9/30,
AIB Data Centers signed a capacity contract with Nebius. This partnership, and others like it, will help build out the needed capacity that Nebius will operate. These contracts help spread the CapEx load so that risk isn't solely on Nebius and the funding isn't through pure debt or dilution.
Nebius has done nothing but affirm my belief in the company. I expect to add more on weakness, despite how much of my portfolio it already makes up.
$MU // Up 15.1% // 18.7% Allocation
Micron is the second largest position in my portfolio for good reason. They are one of three global DRAM makers and the only US based HBM producer, supplying the memory and storage that have become the tightest bottleneck in the entire AI buildout. Their business is rapidly shifting from cyclical commodity to multi-year contracts. I am making a strong bet on them because I don't think the market has fully digested this company's role in the AI buildout and what it means economically. Their valuation is comically low right now, whichever way you look at it; TTM P/S = 9.1, Forward P/S = 4.7, TTM P/E = 14.5, Forward P/E = 6.1. Think about it this way, Micron trades at 6x forward earnings, while projected to out-earn every single Mag-7 company in quarterly operating income, who trade at 25-35x their forward earnings. The gap is the market pricing in a bunch of supply coming in to satiate the demand spike we're in the midst of right now. I simply don't think that supply will have as negative an impact, as quickly as is being projected right now.
Micron had Q4 earnings on 9/30, the last day of the month. They had another blowout quarter. Revenue came in at $54.2B, beating last quarter's guide of $50.0B by 8.5%. Revenue was up 31% QoQ and 379% YoY. Gross Margin (which is the metric to watch for commoditization) was 87% versus the 86% guide. Memory prices continue to increase along with demand. DDR5 prices are up 100%+ in 2026 alone. Forecasts saw another 10-15% increase for DRAM ($39.8B {73% of revenue, +343% YoY}) and 15-20% increase for NAND ($14.1B {+526% YoY}) prices from this quarter to the next one. These prices are not commoditizing yet, they are rising and accelerating. Their Strategic Customer Agreements signed this quarter was 26, up from 16 last quarter. 75% of expected 2027 output is already committed. 2028 output is already being negotiated now. Not only did they knock this quarter out of the park, they also gave impressive guidance. Next quarter revenue guidance is $61.5B, (13% QoQ, 350% YoY), gross margins of 86.3% (which they called the floor for FY27), and said supply-demand will be tighter in 2027 and 2028, than 2026, with no current visibility into when it might balance out again. That is bullish beyond belief in my view. Again, especially at this valuation, these business numbers are a screaming buy.
$NVDA // Up 5.6% // 12.8% Allocation
Nvidia is the leading designer of GPUs, CPUs, networking, and software stacks that constitute most of the world's AI compute. They are the central supplier, investor, and platform layer of the entire AI buildout. The stock has been mostly stagnant for a while, but I am happy holding as the business continues its unprecedented run as the massive leviathan powering the entire emergence of an AI-enabled world.
On 9/2, they agreed to acquire Hugging Face for $12.9B (including a $1B retention program), a distribution hub of open-source AI, where models, datasets, and apps actually live. This helps anchor the open model ecosystem to Nvidia hardware and CUDA by default, widening their moat a little. On 9/9, they expanded their AI infrastructure into a new geography through a partnership that gets Australia's data center ecosystem involved. A day later on 9/10, they released details on a Palantir partnership about sovereign intelligence for critical supply chains. That same day it was reported that Nvidia was reportedly weighing a $10B backing of Anthropic's IPO. This aligns with the circular financing bear case that I discussed earlier this month. Making equity stakes in the labs that buy their own chips is a move that comes with both advantages and risks. Side note, if Anthropic does IPO, I will most definitely look into starting a ~5% position or so.
On 9/17, Nvidia announced a $2B investment into Brookfield's AI fund to help meet upcoming energy and physical demands for the AI buildout. Then later in the month, it was reported that China might allow Alibaba to buy Nvidia RTX chips and on 9/28, a Chinese state-backed firm disclosed a Blackwell chip deal. These are both signs that the export-control wall is cracking from all sides. China is permitting these purchases, while companies are disclosing those purchases openly. Current Nvidia projections exclude China, so any progress in the country is additive to the TAM. Responding to recent safety incidents, on 9/28 they launched the 'Open Agent Safety Platform', to monitor agent behavior, providing a governance layer and smartly trying to monetize AI safety. Also on 9/28, the largest buyback authorization increase in history was announced at $150B, bringing the total remaining to $235B. The fact that they have this much cash to buy stock through FY28 and are still compounding revenue at 100% YoY, while pointing to 70% growth in 2028, is absolutely insane and previews that they believe the stock is mispriced too low right now. I agree.
$RDDT // Down 1.3% // 10.0% Allocation
Reddit is a platform of niche communities whose moat in the AI era is authentic human-generated content and the communities themselves, built and centered around any topic imaginable. 95% advertising revenue, 5% data licensing. They have had impressive revenue metrics with 6 quarters in a row of 60%+ YoY growth, paired with gross margins over 91%, plus a low valuation of 10.2 TTM P/S and 7.3 Forward P/S, which had been cut in half from 20-22x in late 2025. Stock has not responded well recently, but I believe the ticker will catch up to the numbers soon enough and reverse this multiple compression we've seen.
Not too much news this month. On 9/10, Piper Sandler data showed August user growth hit its fastest monthly pace of 2026, a datapoint that indicates improvement in one of the most worrisome metrics of the last few reports (user growth, specifically DAUs and WAUs). Then on 9/22, Reddit got a win in court against Anthropic over data-scraping litigation. It's only a procedural ruling rather than a merits based one, but Reddit's chief legal officer Ben Lee, said "the ruling sends a message that AI companies can't scrape people's information without clear limits on how the data gets used." Apart from the stellar revenue numbers, my main investment thesis on Reddit, is that their data is severely underpriced to the point that they can charge at least 5-10x more for it, probably even more than that. Every win in court like this strengthens Reddit's negotiating position for repricing their data access.
$NET // Up 22.6% // 7.3% Allocation
Cloudflare is positioning themselves as the infrastructure, controls, and payment rail for the agentic internet. They operate a global edge network with CDN, DNS, DDoS, zero trust, and Workers. They are a traditional SaaS plus consumption business. One of my longest held companies. Solid, reliable performer with a sky high valuation (49.5 TTM P/S, 38.1 Forward P/S), which led to some trimming last month.
On 9/2, Cursor, a leading AI company, launched Cursor Cloud Agents on the Cloudflare Sandboxes execution layer. Confirmation that they are taking a part of the agentic workload pie. The day after on 9/3, they partnered with OpenAI on their Daybreak models for AI-powered vulnerability management at the edge. This is their security revenue getting an AI-native upgrade rather than being disrupted by it. Good sign that they are adapting well to the new AI economy.
Biggest announcement of the month came on 9/9, when Cloudflare management announced a raise in their "North Star" growth target from 40% to 50%. That is the exact type of ambitious announcement you want to see. 36% was the YoY number they hit last time, so significant acceleration may be ahead! The same day they disclosed 2M developer adds QoQ (record pace), 80% of major AI enterprise infrastructure companies as customers, and an ongoing shift from subscription to consumption revenue. 'Monetization Gateway' and 'Cloudflare Wallets' introduced as two new products to take advantage of the agent economy and autonomous agent transactions made amongst each other.
There were a few other product announcements. On 9/15, Cloudflare launched new crawler controls that split AI traffic into Search, Agent, and Training categories, with a 'Disallow AI Training' setting, an 'Accountable' crawler designation, and defaults that block training and agent crawlers on ad-supported pages, so multi-purpose bots like Googlebot can be blocked from training even while search indexing stays allowed. This makes Cloudflare the toll booth between publishers and AI companies by breaking the bundling that let training scrapes hide behind search. On 9/25, they disclosed a small cross-tenant data-isolation vulnerability that they quickly fixed in 'Containers' and 'Sandboxes'. Looking at other future long term growth potential, on 9/29 they launched a public 'Certificate Authority' for the post-quantum web, as well as a Deutsche Telekom partnership for a "secure and resilient internet" and European distribution.
$ALAB // Up 28.1% // 6.3% Allocation
Astera Labs is a fabless designer of connectivity silicon for AI racks. Product line includes: Aries PCIe retimers, Taurus Ethernet signal conditioners, Leo CXL memory controllers, and Scorpio fabric switches routing traffic between GPUs, with content-per-accelerator as the growth driver. They've always been pretty expensive valuation wise (50.6 TTM P/S, 23.2 Forward P/S), but as revenue continues to build, the forward metrics actually look more reasonable. Especially with the pre-announcement that Astera expects Scorpio to become its biggest product line a quarter ahead of schedule. On pure revenue acceleration trajectory, this may have the second steepest potential curve, only behind Nebius.
On 9/15, they launched the Leo X-Series, expanding their Leo Smart Memory Controller family for agentic AI and general purpose cloud workloads, pairing with Scorpio X-Series fabric switches to enable memory expansion and pooling. This was explicitly pitched as optimizing 'token economics' as memory footprints explode amid tight supply. Better and more efficient memory is drastically needed to address the current demand bottleneck. When DRAM is the most expensive, scarcest input in the rack, silicon that pools and reuses memory across accelerators stops being a nice-to-have and becomes a cost-of-goods lever for every hyperscaler. Interestingly, this is the other side of the coin when compared to Micron. The memory crisis that powers Micron's pricing is simultaneously creating Leo's market. Astera is helping with this product announcement. Astera is bound to make a ton of money on it too, but it will also allow Micron to sell more volume once they actually have the supply. A massive win-win in my view.
$CRWD // Up 22.8% // 5.5% Allocation
Crowdstrike runs Falcon, a cloud-native cybersecurity platform delivering endpoint, cloud, identity, data, and SIEM protection through a single agent. They are my steady cybersecurity anchor, now pivoting hard to securing AI agents themselves. Very high valuation at 51.4 TTM P/S and 41.7 Forward P/S.
The beginning of the month was headlined by Fal.Con 2026, their yearly conference which included a dozen plus announcements, 10,000+ attendees, and 4,000+ organizations. Some important announcements include: Falcon Guardian, their headline product for discovering and securing AI agents across an enterprise. It's an entirely new module that is attempting to create the first solution to 'agentic security' right as the need for it is arising. Impressive agility for a large company and why they are a leader in their field. ; Agentic IdP, which gives AI agents trusted identities through the Falcon Identity control plane. ; SafeMind, announced with Jensen Huang on stage, a purpose-built agentic AI system for cybersecurity that combines offensive and defensive models in a continuous learning loop. ; Google Cloud AI-security expansion with a 'Cyber Superintelligence Lab'. ; an EY.ai partnership to make Falcon their enterprise AI security foundational blueprint. ; and an expanded OpenAI partnership to ensure Codex agents and GPT-5.6's cyber model comes to Falcon.
At Citi's Global TMT Conference on 9/10, Crowdstrike CFO Podbere argued AI is reshaping security spending in their favor with stronger demand, faster module adoption, while acknowledging margin pressure and Microsoft competition. On 9/17 and 9/24, they were named a leader in Forrester 'External Threat Intelligence' and 'Proactive Security Platforms', ranked highest in Strategy. On 9/28, their Nvidia partnership expanded to cover the entire AI stack, securing a seat inside Nvidia's agent-security architecture. Then on 9/29, Falcon was officially listed on the OpenAI Marketplace. Falcon is also already on Claude Marketplace.
$SNOW // Up 7.1% // 3.6% Allocation
Snowflake is an AI Data Cloud with consumption-priced storage, compute, and data sharing across AWS/Azure/GCP, now layered with Cortex, the CoWork knowledge-worker agent, and the CoCo coding agent, positioning itself as the governed control plane for enterprise AI. They are a durable, steady compounder that is reaccelerating at scale and currently has a historically decent valuation (22.1 TTM P/S, 16.8 Forward P/S).
Big news came early for Snowflake as they had their Q2 earnings on 9/2 at the beginning of the month. Product revenue was $1.49B (up 37% YoY), beating last quarter's guidance of $1.42B by 5.2%. While those aren't the fastest numbers we've seen, it was the third consecutive quarter of product revenue growth acceleration. CoWork and CoCo, were given special shoutouts as they led a meaningful step up in AI revenue with new workloads and consumption. Their product revenue guide for FY27 was raised from $5.84B (31% growth) --> $6.07B (36% growth) and would be a 3.9% full-year raise. Very big considering their current scale and usually minuscule raises. NRR slightly increased from 125% --> 126%, 828 customers spending over $1M (up 27% YoY), and 829 of the Forbes Global 2000 as customers. RPO grew 30% YoY to $9.0B, but declined sequentially from $9.21B (38% YoY).
Macro:
In last month's macro section I spent a lot of time discussing some theories and possibilities surrounding the future of AI and its impact on the economy and the world more generally. TLDR; there are just as many reasons to be fearful and cautious as there are to be excited. This month I wanted to shift the conversation back to the K shaped economy exacerbating wealth inequality through asset ownership, stagflation, and the massive potential energy behind the stocks in my portfolio.
Starting with the K shaped economy. I am increasingly of the opinion that wealth inequality, standards of living, and the general gap between the haves and have nots will continue to get much worse before it gets better (at current trajectory). I see a world where consumer wage growth tops out at 3-4% per year, while assets compound at 50-100%. AI is structurally shifting income from labor to capital, which is already the direction money normally flows. The productivity gains go to whomever owns the compute, memory, and models.
One overall macro question I had during my bear case analysis was about who sustains demand for these AI products? My fear is about a weak consumer base that can't support the infrastructure we're currently building. Right now, corporations are buying from corporations. The $730B hyperscaler capex buildout, the prepayments for Nebius and Micron, commitments from Nvidia, etc. It's a closed capital loop that, right now, doesn't need healthy consumers yet. There are a few ways this closed loop might open up:
AI productivity appears in the real economy. GDP, margins, and business metrics go up and show true gains that validate the CapEx spending, and in turn, leads to more spend.
Consumer side cracks. If real wages continue to trend negative, eventually the worker economy will bleed into the AI economy and spend will decrease. Advertising is one way, enterprise software is another.
Political/government intervention. A world where companies continue to hit record metrics, while paychecks shrink in real terms is not politically nor mathematically stable. Release valves can be done through either redistribution (capital gains, windfall taxes, AI taxes, antitrust) or it can be done through populist rupture and upheaval of current systems/institutions.
I see each of these happening to some extent. Hopefully we can address some of the root causes sooner rather than later...
The next topic is stagflation. The current economic landscape shows some striking similarities to what the US dealt with during the 1970s. Stagnant growth (other than AI) + persistent inflation + rising unemployment + supply shocks. We've had inflation above target rates for 5 years in a row now. The recent bout of inflation being exacerbated by tariffs and war-driven energy costs, which followed immediately after a successful soft-landing that balanced rate hikes and unemployment from 2022-2024. Self-induced pain on inflation. Payrolls decelerated, unemployment ticked up very slightly to 4.2% (10/2 event), wage growth is at a five year low and negative in real terms. The Fed is facing the same trap from the 1970s: hike rates and you risk damaging an already weakening labor market; hold or lower rates and you make inflation worse. I worry that stagflation occurs for labor, while the asset economy inflates, again all going back to the K shaped economy described above.
Inflation was unchanged from July. August CPI up 0.1% MoM, 3.4% YoY for headline inflation. Hotter than expected leading to the Fed's decision to hike rates by 25bps to 3.75%-4.00%. This follows an August jobs report of +133k, above forecasts and reversing from a negative month in July (but recent 10/2 report has worse numbers). It was accompanied by a statement saying the rate increase supports a "timelier return to 2%." The impact on my stocks could be negligible or it could massively cut valuations like in 2022. I do not anticipate rate increases or decreases to impact how I invest.
To conclude the macro section I wanted to briefly discuss a 'vibe' I'm sensing right now. This is purely intuition and gut feeling, but it seems like my stocks (AI mostly) are coiling. This may sound greedy considering the gains that I've seen this year already, but I can't help but feel 1) that the overall macro economy is holding my portfolio back and that valuations are slightly compressed because of it and 2) growth will be a lot more explosive and last longer than even the rosiest estimates are taking into account.
Not only are current valuations pretty low, but I also think 2028 and beyond are mispriced due to the market's current bet that supply will flood and normalize demand by then. I just don't see that in the numbers right now. I see a massive economic disruption that has only just begun. I sense that Mr. Market is just waiting for the right macroeconomic conditions to massively rip upward.
Final Portfolio (110% Long):
$NBIS 45.8% π
$MU 18.7% π
$NVDA 12.8%
$RDDT 10.0%
$NET 7.3% π
$ALAB 6.3% π
$CRWD 5.5% π
$SNOW 3.6%