Nova AIOS vs Palantir Foundry: a founder's side-by-side

I have spent the last two years building Nova AIOS, an AI operating system designed for Philippine and Southeast Asian enterprises. This comparison draws on close study of Palantir Foundry's architecture, pricing, and published deployments, and on representative procurement scenarios for SEA mid-market lenders where a Foundry quote and the math behind it tend not to work. The figures and scenarios below are illustrative, not reports from signed customers.

This is not a takedown of Foundry. Foundry is the best AI operating system in the world for a specific shape of customer. The reason Nova AIOS exists is that almost no enterprise in the Philippines or Southeast Asia is that shape, and a $500,000 USD floor price will not become a $50,000 USD floor price no matter how much we wish.

The honest version of the comparison is structural. Foundry and Nova share the ontology-plus-agents pattern, but they optimize for opposite ends of a market. This post lays out the differences in a way you can actually use to make a decision.

Palantir Foundry is the right call if your enterprise has more than ten thousand internal users, operates across at least five regulators in different countries, requires FedRAMP High or DoD IL5 certification, or has a procurement department that will not sign with a vendor whose stock does not trade publicly. If any one of those four applies, stop reading and go talk to Palantir.

Nova AIOS is the right call if your enterprise is mid-market by Philippine or SEA standards, runs on a budget below $200,000 USD per year, needs RA 10173 and BSP regulations baked into the permission layer rather than bolted on after, and would benefit from agents that speak Tagalog or Cebuano natively. The rest of this post is the longer version of that paragraph.

This is the table I wish someone had handed me in 2023.

Read that table twice. Almost every row reads as "Foundry has more" or "Foundry has bigger." That is correct. The interesting question is which of those mores you actually need. Most Philippine enterprises do not need any of them, and pay an order-of-magnitude premium when they pretend otherwise.

There are real Foundry-only deployments. I am not going to pretend otherwise. Four shapes.

• Multi-country regulated rollouts. If your firm operates in fifteen countries each with its own data sovereignty rules, Foundry has been doing that work since the GDPR rollout. Nova can be ASEAN-aware. We are not ready to be everywhere-aware.
• Defense and classified-adjacent work. Foundry holds security certifications a Filipino startup will never get and probably should not try to. If your customer is a defense ministry or a national-security-critical operator, choose Foundry.
• Ten-thousand-seat internal rollouts. Foundry can scale an ontology and a permission model across a Fortune 100 with deep internal IT. Nova is built for the mid-market shape, typically below 2,000 internal users.
• Procurement departments that require a publicly listed vendor. Some banks and government bodies have an absolute rule that revenue-critical software must come from a vendor with audited public financials. Palantir trades on the NYSE. Nova does not.

If you fit any of the four shapes above, this post is making the wrong recommendation. Keep Foundry on the shortlist and treat Nova as a vendor for the smaller, more local layer of your AI stack.

Five places. None of these are theoretical. All of them come from actual customer evaluations where the prospect went with Nova after pricing Foundry.

• Regulation baked into the permission model, not bolted on. Nova ships with RA 10173, RA 11765, and the SEC MC 18 (s.2019) collection-conduct rules as opinionated defaults, and respects the BSP Circular 1133 rate and fee ceilings on covered loans. If a collections agent tries to contact a borrower outside the permitted calling hours or through a prohibited third party, Nova refuses. These rules are in the codebase, not in a 60-page policy PDF your engineers have to remember.
• Tagalog and Cebuano models that understand the country. Nova fine-tunes its LLM layer on Philippine call-center transcripts, BSP examiner reports, and barangay-level government documents. The difference is visible the first time you feed a real customer escalation into both systems.
• Twelve weeks instead of twelve months. A Foundry deployment with a US integrator runs twelve to eighteen months to first production agent. A Nova deployment runs eight to twelve weeks. For a Philippine MFI that needs an agent in front of customers this quarter, that gap is not a luxury.
• Founder-accessible support. A Nova customer talks to me directly, on Viber, within four hours during Manila business hours. Not a sustainable model at a hundred customers, but for the first thirty it is the difference between an AI vendor that knows your business and one reviewing your ticket.
• Pricing that survives a Philippine CFO conversation. A bank's CFO will not approve a ₱28M-per-year platform for a tool it could replicate, badly, with three engineers at ₱9M a year. The Foundry math does not pencil out at the median PH enterprise size. Nova's does.

This is the section procurement teams care about. Three-year total cost of ownership for an equivalent deployment serving a Philippine enterprise of roughly 1,500 employees, with three AI-driven workflows: customer onboarding, collections triage, and an internal operations dashboard.

The gap is approximately 16x. That ratio holds across the deployments I have priced for actual Philippine enterprises. The mistake some teams make is assuming the ratio reflects quality, that Foundry must be sixteen times better. It is almost entirely engineering-economics arbitrage. A senior Filipino AI engineer at Nova costs roughly $30,000 USD per year fully loaded. A senior Foundry forward-deployed engineer based in Singapore costs roughly $180,000 USD per year. The price ratio in the deployment reflects the price ratio in the labor underneath it.

You are paying for the same architectural pattern. You are paying very different prices for the humans who build and maintain it.

Deployment A, a 12,000-person multinational bank with operations in PH, SG, HK, and AU, pays for Foundry. The right call. They need cross-jurisdictional data residency, a procurement department that will not sign with a private LLC, and SOC 2 Type II with annual third-party audits. With a $4M USD annual AI budget, the Foundry quote does not blow up the plan. Nova sits on their shortlist as a satellite vendor for Philippine-specific add-ons.

Deployment B, a Cebu-based microfinance institution with 240 staff and 41,000 borrowers, pays for Nova. The right call. Their annual technology budget is ₱18M. A Foundry contract would eat it for two years. They need RA 11765 and SEC MC 18 collection-conduct compliance, which Foundry would require them to configure manually, and their use cases are bounded: collections triage, KYC, customer service. Nova ships in ten weeks for ₱950K plus a ₱120K monthly retainer. The same workload on Foundry would cost roughly thirteen times more for capabilities the MFI does not need.

Deployment C, a Philippine national government agency with critical-infrastructure responsibilities, is messy. The agency wants Foundry's security profile, Nova's price, and a local vendor's procurement profile. The honest recommendation is a split: pay Palantir for the critical-infrastructure slice where the certification stack is non-negotiable, and pay Nova for the citizen-facing layer where Tagalog-native models, RA 10173 defaults, and twelve-week delivery matter more than the FedRAMP badge. Even there, the dollar split is roughly 80 percent Foundry and 20 percent Nova.

If you are running a head-to-head right now, three things will help you decide faster.

• Get both prices in PHP. Foundry quotes in USD. Have your CFO convert and present the three-year TCO in pesos. The conversation changes when the number is ₱125M instead of $2.25M.
• Bring a real workflow to both vendors. Pick one actual business process, ideally one that touches a regulated activity, and ask both to scope it. Foundry will produce a sophisticated proposal. Nova will produce a shorter, faster, cheaper one. The right vendor reveals itself in how each scope reads against your constraints.
• Ask both which workloads they would refuse. Foundry will tell you they do not take on single-workflow deployments under a half-million dollars. Nova will tell you we do not take on multi-jurisdictional or classified work. The presence of refusal is the signal of seriousness.

Foundry is the best AI operating system in the world for the customer it was built for. Nova AIOS is the best AI operating system in the Philippines and Southeast Asia for the customer Foundry was not built for. These are not contradictory statements. They describe a market that has split, the same way databases split into Oracle for the Fortune 500 and Postgres for everyone else.

If you are a PH or SEA enterprise CTO, CFO, or procurement head making a real decision, the highest-leverage move is thirty minutes with both vendors in the same week with the same workflow in hand. The right answer will be obvious by the end of the second call. The Nova half of that comparison is on the house.