Going into the latest earnings, PLTR saw a steep eight-month correction, falling by 50% from peak to trough. This is all while Michael Burry openly voiced his shorting (or put-buying) of the stock.

Meanwhile, our Multi-model Signal sat out of PLTR in late February until the beginning of August.

A review of the individual models suggests that the ML model was particularly sensitive to the initial decline, stepping away almost perfectly at the peak. The Neural model was also mostly Risk-off during the decline, but with several attempts to play the counter-trend rally.

Trend Model

The deployed trend model uses the 50-day vs 200-day moving averages to filter trends. And it’s a “lenient version”, meaning we only step out when the price and the 50-day MA are both below the 200-day MA. It’s “lenient” because the default is to keep the PLTR position long, which has historically worked out better than the other way round.

Regime Model

Our Regime model flashed uncertainties as PLTR peaked, marking the price regime as “Sideways” (yellow) and “Bearish” (red). In terms of positioning, our MS actually stay in position during both the Bullish and Sideways regimes.

While that sounds aggressive, the net result is that the Regime model has “over” captured PLTR’s mesmerising rise in the past 2 years, and the model alone would have beaten buy-and-hold by 102%.

For a stock like PLTR that rises steeply with enormous valuations, I think obsessing over the small-picture details could make you miss the big tree. If one just follows the overall MS strategy with reasonable position sizing and ignores all the other noise, they would still have beaten the S&P 500 by 10-fold without much stress.

Palantir sells data-integration and AI operations software to governments and large enterprises. It runs two businesses that behave nothing alike, and the split is the whole company.

The government half is contract-driven, multi-year, slow to win and very slow to lose. Programmes run through accreditation, security clearances and installation inside isolated networks, which is why a Program of Record win compounds. US government revenue was $809m in the June quarter, growing 90% YoY.

The commercial half runs on land-and-expand: a small paid pilot, then an expansion once the customer's own data has been mapped into the ontology. US commercial revenue was $764m, growing 149% YoY and 28% QoQ, and it accelerated again rather than fading.

The geography matters more than the segment. US revenue of $1,573m grew 115% YoY and is now over 81% of the company. International commercial revenue was $182m, growing 26% YoY and just 2% QoQ, which is effectively flat.

This means either the commercial proposition does not travel, or it has not travelled yet. That single comparison is the most important unresolved question in the business, because it decides whether there is unconditioned AI demand signal or simply a domestic AI budget surge.

Net dollar retention reached 157%, up 700bps QoQ.

Upstream: Palantir sits at the top of the stack, above the data and above the compute. Underneath it are the cloud platforms it deploys on — Microsoft (MSFT), Amazon (AMZN) and Alphabet (GOOGL) — plus their government-accredited regions.

One step further down is the silicon, where a partnership announced in late June 2026 puts Nvidia (NVDA) open Nemotron models and accelerated computing inside the platform for classified and isolated deployments.

Downstream: US federal agencies, which are roughly 42% of June-quarter revenue, allied governments, and a commercial base expanding through the land-and-expand motion.

Peers that matter: Snowflake (SNOW), MongoDB (MDB) and Databricks (private) store and process the records Palantir models. That relationship is two-sided because many Foundry installations sit on top of one of these vendors, but they also build the semantic layer Palantir sells. Palantir’s closest rivals for the enterprise budget are ServiceNow (NOW), Salesforce (CRM) and Oracle (ORCL).

Where the cycle sits. Enterprise AI software is early in an adoption cycle but Palantir may be late in a valuation cycle.

Corporate AI budgets are still expanding, and Palantir's own numbers accelerated in the June quarter. However, Palantir’s valuation has nosedived, with its share price tumbling by half between Q4 2025 and Q2 2026. This shows that the market is happy to punish high-valuation companies when the sentiment turns.

Revenue grew by 93% YoY in Q2 to $1.9 billion, the highest reported growth rate in the company's history.

Management framed the quarter as the arrival of a major shift (Alex Karp has been vocal about this in the news. Enterprises are rejecting a token-consumption model and buying what they call AI sovereignty, meaning ownership of data, logic, weights and audit within their own boundary.

Remaining performance obligations reached $4.9 billion, up 103% YoY, and total remaining deal value $13 billion, up 83% YoY. US commercial total contract value bookings of $2.1 billion 153% YoY. Backlog is growing faster than revenue, which means growth can be sustained in the near future.

Guidance:

Third-quarter revenue is guided to $2.16 billion (11.7% QoQ). Adjusted operating income is guided to $1,292m to $1,296m, a margin near 60%.

The full-year revenue guide rose to $8.15 billion, or +82% YoY. The full-year US commercial revenue is guided to at least $3.4 billion, a minimum of +134% YoY, which is a deceleration compared with the 149% YoY delivered in the June quarter.

One should also note that the “land” motion (in “land-and-expand” is decelerating even as the expand motion runs hot. US commercial customer adds went +9%, +8%, +8%, +6% QoQ across the last four quarters. Total customers grew +24% YoY and only +4% QoQ. But net dollar retention of 157% means the entire acceleration is coming from spending more within an existing base of a few hundred US commercial accounts.

Pricing power is real and visible in the margins. Gross margin is 85%, while the free cash flow margin is 62%. But stock-based compensation ran at 13.6% of revenue over the last twelve months, so part of that cash margin is paid for in shares rather than dollars.

Palantir’s SBC levels are very high, although still below its revenue growth.

Here are the key financial metrics from the past four quarters. I designed the table this way so we can focus solely on the key top-line, profit, and cash flow metrics.

Here are Wall Street's forward estimates.

PLTR’s most relevant valuation multiple is EV/ Sales, which in 2025 reached an eye-popping 150x (stock price around $200).

However, since Q4 2025, valuation has halved, as the stock price sits 15% below the ATH despite the 79% LTM revenue growth. At 71x, PLTR’s EV/ Sales trades near its 3-year median.

Today’s EV/ Sales is actually pretty “fair” compared with PLTR’s historical regression between valuation and growth.

I have assigned a NTM EV/ Sales multiple of 60x as the price target, given PLTR’s 58% next-3-year CAGR. Again this is in line with the historical norm.

This implies a price of $258, about 47% higher from here.

The DCF analysis below assumes that PLTR’s revenue growth slows to about 40% in 5 years and steadily declines to a terminal growth rate of about 3%. It also assumes a stable EBIT margin around 63%, with very little change in D&A, CapEx, and networking capital. The WACC assumes that the long-term capitual structure carries 20% debt.

The DCF “fair value” ist at $235 given the above assumptions, or 34% above last night’s closing price.

Investors are paying for a near-perfectly executed Enterprise AI business with incredible growth and an entrenched US Government moat. The valuation looks punchy by a normal company standard, but is justifiable given the above.

However, PLTR’s weaknesses will come from the Commercial arm, while international growth doesn’t show the same enthusiasm either. My suspicion is that PLTR’s deep involvement with the US government is a double-edged sword, as other Sovereign buyers may be concerned about their data security.

PLTR rose from $125 pre-earnings to $174 today, in just a month. That jump has cleared the moving-average cluster and the original resistance region near $160. The bottom of the Q4-Q2 selloff is most probably in, and it’s a question of whether it retests the ATH before the next pullback.

I would lean towards yes - the momentum is on its side with the broader software sector recovery.

My overall read is that, on a discretionary basis, I would normally find it difficult to pull the trigger after such a steep run. However, the valuation is still “reasonable” (as reasonable as a name of such high growth and free cash flows can go), while the momentum is still strong. I would be a happy buyer should PLTR get picked by the Systematic Portfolio.