Sementis Insights | AI Is Exponential. The Harvest Is Not.

This was the week Sequoia decided that menos es más and announced it is closing its London office, while Revolut was valued at US$ 115 billion: some VCs are slimming down, some fintechs keep growing. Out in the field, Vereda sells farm inputs over WhatsApp and Celes left Colombia with its eyes on Brazil. Anthropic opened its S-1, which guarantees weekend reading. And physical AI, which dreams of robots in the field, is still negotiating with the internet signal. Grab your coffee, open the links and let's go: today we separate what is fact from what is "sources say." Buen provecho!

But first, subscribe to the newsletter for free to receive it every Friday:

And if you want to follow the daily news on WhatsApp, follow the channel!

1. Fintech 100: four Latin American companies among the 100 most promising. On Oct 28, CB Insights released the 9th edition of the ranking: 100 winners across 21 categories, picked from more than 70,000 fintechs from pre-seed to Series C. The region appears with Aviva (consumer lending), Tapi (cross-border payments), Clara (business banking) and Pluggy (embedded finance). The report highlights three themes: blockchain in mainstream financial services, financial institutions' own data as an underused AI asset, and regulatory burden turned into product. Why it matters (our analysis): only one of the six market-map sections is consumer-facing, and three of the four Latin American picks work on infrastructure or services for businesses. Sources: CB Insights and Startups.com.br (Oct 28). Descriptions of the Latin American companies come from Startups.com.br.

2. Women in Tech & Capital: more exits in the US, with caveats. According to PitchBook ("US All In 2025"), via Startups.com.br, 24.9% of US VC exits in 2025 involved companies with at least one female founder, up from 23.6% in 2024; that was 349 exits versus 290, and exit value reached US$ 51.1 billion (18.4% of the total), versus US$ 20.1 billion (13.3%). In PitchBook's release (Business Wire), female founders raised a record US$ 73.6 billion in 2025, with about two-thirds of the dollars going to AI and more than US$ 30 billion coming from Scale AI and Anthropic. Lombard Odier's LO Women Invest study (September 2026) restates these data and adds average US VC pre-money valuations: US$ 33.3 million for companies with at least one female founder, US$ 39 million for all, and US$ 19 million for all-female-founded firms. Caveats: exit value includes extrapolated figures, the sample is VC-backed companies with at least one female founder (even if she has left), and the numbers are for the US. In Brazil, ABStartups' 2025 Startup Ecosystem Mapping shows cis women at 19.9% of founders (self-reported sample of 3,650 startups), a composition figure, not a capital figure. We found no reliable, current source on capital going to women founders in Brazil.

3. Physical AI and agriculture: where the bottlenecks are. According to Bloomberg Línea (Nov 4), agriculture is the main driver of physical AI among Brazilian startups. The AWS report "AI Gets Physical" (October 2026) surveyed 4,420 startups in 20 countries, and Brazil appears as an emerging market. Globally, 70% of builders say slow customer adoption is constraining their growth and 26% say their economy is ready to scale. More detail and our take in The Bigger Picture, just below.

4. Anthropic: S-1/A says it is close to "RSI." According to TechCrunch (Nov 2), which reviewed the S-1's risk-factor section, internal estimates indicate the company may soon reach the Responsible Scaling Policy's definition of dramatically accelerated AI-driven progress; no current model is above the threshold. Why it matters: a regulatory disclosure relevant to anyone assessing AI risk and governance. Secondhand reading: we did not open the original document.

5. Global VC reshapes itself; Revolut becomes a benchmark. Sequoia announced the merger of its early-stage and growth teams and the closing of its London office and its Arc and Scouts initiatives, according to TechCrunch (Nov 4). Revolut was valued at US$ 115 billion in a secondary sale, about 53% above the US$ 75 billion of November 2025, according to Startups.com.br (Nov 4), which lists Brazil among future markets, with no timeline. Why it matters: less presence of global VCs outside the main hubs opens room for local funds; Revolut's valuation is a reference for fintechs in the region.

6. Bayer and IdeeLab: biofungicides for Brazil. Bayer's agricultural division and Brazilian biotech IdeeLab will develop biofungicides for tropical pests in a 12-month partnership, with three candidates already in the field, according to AgFunder News (Nov 4). Bayer said it will launch 17 new biologicals in Latin America by 2030. Why it matters: a sign that biologicals R&D is being outsourced in Brazil.

The Bigger Picture — By Juliana Chini

Brazil has the potential to lead applications of physical AI in agribusiness. But is our investment model prepared for the time that innovation in the field requires?

Throughout my career in the agribusiness innovation ecosystem, I have met startups with extraordinary technologies. Solutions able to reduce waste, increase productivity, optimize operations and solve problems that have followed rural producers for decades.

Some thrived. Others, despite their technological potential, could not move forward.

And, often, the problem was not the technology. It was the math that did not add up.

A report published this week by Bloomberg Línea, based on an AWS survey, brought that reflection back. Artificial intelligence is moving beyond screens and into the physical world, through drones, sensors, robotics, computer vision and smart equipment.

And Brazilian agribusiness is emerging as one of the most promising environments for that transformation.

According to the report AI Gets Physical: The Global Startups Readiness Index, which surveyed 4,420 startups in 20 countries, 15% already develop physical AI solutions. In Brazil, that share is 12%, according to data released by Bloomberg Línea.

Agriculture plays an important role in this story. In the survey, 47% of Brazilian startups point to agriculture among the first sectors that physical AI should transform, versus 25% in the global sample.

It is not hard to see why.

We have large-scale agricultural operations, logistical challenges, pressure for productivity, a shortage of specialized labor in certain activities and a permanent need to produce more while making better use of available resources.

And we have companies developing concrete solutions.

BemAgro, for example, uses artificial intelligence, computer vision and geospatial data to turn images from drones, satellites and machines into information applicable to agricultural planning, monitoring and execution.

IA SENSE, from Sertãozinho, develops computer vision solutions integrated with cameras, sensors and devices, with applications that include crop monitoring and industrial operations.

They are examples of how artificial intelligence can leave the purely digital environment and contribute to decisions and processes in the physical world.

But there is an enormous difference between developing a solution that works and building a company capable of scaling it.

In agriculture, the pace of technology does not always match the pace of capital.

Imagine a startup that develops a solution to identify planting failures or optimize a specific agricultural operation.

It needs to find a customer willing to test it, negotiate the pilot, integrate equipment and data, work within a specific operational window and demonstrate results.

If the test depends on a harvest, the next validation opportunity may be months away.

If the first pilot needs adjustments, it may be necessary to wait for another crop cycle.

Meanwhile, the startup keeps paying salaries, developing technology, sustaining infrastructure and burning cash.

The investor, in turn, wants growth, commercial traction and evidence of return.

And the producer, rightly, wants to be sure the technology will deliver results before committing their operation.

Everyone has legitimate reasons. But the clocks are not synchronized.

I have seen this mismatch up close. Startups that needed one more harvest to prove results but had no capital to get there. Technically successful pilots that never turned into contracts. Promising technologies pressured by growth expectations incompatible with their adoption cycles.

This is precisely the bottleneck that the AWS numbers help to size.

Globally, 70% of physical AI builders say slow customer adoption constrains their growth. Only 26% consider their economies ready to deploy these technologies at scale.

Access to capital also appears as an important obstacle: only 46% assess that they have adequate financing conditions to develop and scale their solutions.

We are not talking only about a Brazilian difficulty. We are facing a structural challenge of a market that must combine software, hardware, infrastructure, operational validation and behavior change.

And that requires rethinking not only the technology, but also how we finance and buy innovation.

Does it make sense to value an AgTech that depends on crop cycles with the same growth expectations as a corporate software startup?

Could large companies, cooperatives and input suppliers take a more active role in contracting and financing pilots, rather than just offering testing environments?

And are our venture capital, corporate venture capital and innovation financing instruments prepared to sustain the path between a promising prototype and a recurring commercial operation?

Brazil appears in the AWS study as an emerging market in physical AI, while Mexico is already in the group of accelerating markets.

We have agricultural knowledge, potential demand, entrepreneurial capacity and real problems that justify developing these technologies.

But an application advantage does not automatically become market leadership.

The next frontier of agricultural innovation may not be developing more technology. It may be building the conditions for good technologies to survive until they reach scale.

Because, in the end, there is no point in having startups able to revolutionize the field if they cannot survive long enough to prove that the revolution works.

And the question remains: are we financing agricultural innovation, or just demanding that it grow at the wrong pace?

The Bigger Picture — By Juliana Chini

Sementis | Intelligence for What's Next.

Celes (Colombia) raises US$ 5 million with SP Ventures to enter Brazil. A US$ 5 million Series A led by SP Ventures, with Fulcrum Global Capital, 30N Ventures and Taram Capital, confirmed by Startups.com.br (Oct 30), Forbes Colombia (Oct 27) and Forbes Chile (Oct 28). The AI software does demand forecasting and replenishment for retail; the funds will finance a formal entry into Brazil, which is not yet completed. Previous round: a US$ 3 million seed led by FEMSA Ventures in 2024. Valuation and revenue were not disclosed. The metrics (8,500+ stores in 8 countries, 55 million predictions per day) are self-reported. Analysis: a Brazilian fund leading the round of a Latin American startup shows local capital financing the entry of Latin American companies into Brazil.

Vereda raises R$ 30 million to connect input manufacturers directly to growers. According to Startups.com.br (Nov 3), a R$ 30 million seed co-led by Canary and Valor Capital Group, with participation from Speedrun (a16z). The Speedrun page lists Vereda (founded in 2025, Rio Verde, Goiás), without a round amount. The company acts as the commercial arm of input manufacturers, on commission and with no inventory, using a field team and an AI agent on WhatsApp (website). There is a financing pilot in the 2026/27 off-season corn window with financial institutions; Vereda says it does not extend credit or take on the risk. Valuation, revenue, GMV and customers were not disclosed, and the price reduction for growers is a projection of the model. The amount and investors come from a single source. Analysis: direct sales remove layers from input distribution, and whoever finances defines the risk.

Rivio extends its Series A to R$ 180 million. An R$ 80 million extension led by Valor Capital, with Monashees, Positive Ventures and APL, according to Startups.com.br (Nov 4). The healthtech uses AI agents to audit hospital billing; its valuation went from just over R$ 500 million in January to more than R$ 1 billion, with 13 hospital groups as customers. Revenue not disclosed.

Abroad: Trajectory raised US$ 45 million in a Series A, with Sequoia, at a US$ 300 million valuation (TechCrunch, Nov 2). OpenAI-backed Thrive Holdings acquired ADT for US$ 15.3 billion in enterprise value (TechCrunch, Oct 29). Meta confirmed the acquisition of Conductor, a coding-agents startup, without disclosing terms (TechCrunch, Nov 4).

In the next 7 days: Web Summit Lisbon, November 9 to 12. Tech, AI, startups and VC. Tickets on sale; availability not stated. websummit.com

Executive education: MIT Sloan, AI Executive Academy (Cambridge, US, Jan 18-29, 2027, US$ 24,500; round 2 applications due Nov 30, executive.mit.edu) and Oxford, AI Strategy and Leadership Programme (cohorts on Jan 18 and Mar 8, 2027; deadline and availability not stated on the page, sbs.ox.ac.uk).

"The demo and the harvest"

Physical AI: the demo is easy; the harvest calls the shots.

See you next Friday. Have a great weekend!