How many people have a brain chip? Every implant counted, 1998-2026

01 Key findings

1. At least 138 people have had a long-term brain-computer interface implanted, 1998 to October 2026: 67 in trials to December 2023 and 71 announced since.Patrick-Krueger et al. 2024; company, hospital and registry reports; count by IQ Metrics
2. As many in under three years as in the 26 years before: 69 people from January 2024 to October 2026, against 69 from 1998 to 2023, about 24 a year against 3.IQ Metrics count
3. China now implants the most: 45 people since 2024 in NEO, Beinao-1 and StairMed, against 21 for Neuralink. NEO alone has 37, the most of any programme.ClinicalTrials.gov NCT06990412; Tsinghua; BiopharmaTrend; Reuters; Pulse 2.0
4. The first long-term brain implant approved for sale is NEO, approved in China in March 2026 for adults with paralysis from a spinal cord injury; the first paid implant followed on 13 July 2026.BiopharmaTrend; Asia Economy
5. Speech implants reach 78 words a minute, about half the speed of natural conversation (160).Metzger et al. 2023; Willett et al. 2023
6. None of the 47 registered implant trials tries to make anyone smarter: none accepts healthy volunteers and none uses an IQ test.ClinicalTrials.gov, analysis by IQ Metrics

02 Statistics to quote

03 How many people have a brain chip?

At least 138 people have had a brain-computer interface implanted for long-term use: a device that reads signals from the brain so the person can move a cursor, type, speak through a computer or move a paralysed limb. The count runs from the first long-term implant in 1998 to October 2026. A published review found every implanted participant in trials to December 2023, 67 people in 21 research groups ^[1]; it names each one by the code the trials use ^[2]. We added every later implant we could tie to a named source: 71 more.

Figure 1

Every person with a long-term brain implant, 1998 to October 2026

One square per person we can count (138). Neuralink as of 28 January 2026, Beinao-1 as of 20 March 2026, NEO to 13 July 2026. Later, less exact reports would add about 28 more (Neuralink and Beinao-1). Hover a square for the programme, country and device.

University and hospital trials 59Synchron 10Neuralink 21China's new programmes (NEO, Beinao-1, StairMed) 47Paradromics 1

138 peopleone square = one person

The true number is higher, and we cannot say by how much. University trials such as BrainGate keep implanting people, but recent papers leave out implant dates to protect privacy ^[1], so only the implants announced with a date are in our count. Two programmes have given larger but vaguer figures: Beinao-1 “nearly 30” in June 2026 ^[15] and Neuralink a 26th and 27th participant in mid-2026 posts we could not confirm. With those, the total is about 166. We also leave out at least 38 people who had a brain-reading array placed for minutes or weeks during other brain surgery, such as Precision Neuroscience’s 37 ^[21].

04 Is the number of people with brain chips growing?

Fast. Before 2014 there were 16 implants in 16 years, all in four university groups. From 2014 to 2023 there were 53. Then, from January 2024 to October 2026, 69: as many in under three years as in the 26 years before, about 24 a year.

Figure 2

People who got a brain implant in each period

The periods are not the same length (16 years, 10 years, then 2 years and 10 months), which is the point: about 1 implant a year before 2014, 5 a year from 2014 to 2023 and 24 a year since January 2024.

University and hospital trialsSynchronNeuralinkChina's new programmes (NEO, Beinao-1, StairMed)Paradromics

161998-2013

532014-2023

692024-Oct 2026

Neuralink went from its first patient in January 2024 to 21 participants two years later ^[12]. It reported 12 in September 2025, so it implanted about one person every two weeks in the four to five months that followed.

Figure 3

Neuralink: people with its implant, at each count the company gave

Counts as announced, by month of the announcement or implant. Later blog and news posts mention a 26th and 27th participant (June and July 2026); we found no company or wire-service statement to confirm them.

1Jan 2024

2Jul 2024

3Jan 2025

12Sep 2025

21Jan 2026

05 Has China overtaken the US in brain implants?

In new implants, yes. Since January 2024, three Chinese programmes have implanted at least 45 people: NEO from Tsinghua University and the company Neuracle (35 of its 37 since 2024), Beinao-1 from NeuCyber (7) and StairMed (3). Neuralink, the largest US programme, had 21 by January 2026. NEO’s 32-person trial ran from May to November 2025 ^[4], and in March 2026 China approved NEO for sale, the first market approval of a long-term brain-reading implant that we know of ^[6]. The first paid implant followed on 13 July 2026 ^[7].

Over the whole count the US still leads: 51 of the people implanted to 2023 were in the US, plus most of Neuralink’s 21 (the company has not said how many are in Canada, the UK and the UAE). China has 48 in all. Of the 47 implant trials we found on ClinicalTrials.gov, 26 have a start date in 2024 or later (some not yet recruiting), 8 of them in China ^[3].

06 Who has implanted the most people?

NEO (37), Neuralink (21) and BrainGate (16), the US university network that implanted the first Utah array in 2004 ^[1]. Then Synchron (10), whose electrode goes in through a blood vessel instead of open-skull surgery ^[28], and the Pittsburgh and Chicago group (9). 14 programmes have implanted just one person.

Figure 4

Who has implanted the most people

People counted per programme, 1998 to October 2026. 14 more programmes have implanted 1 person each. Beinao-1 would rank second if its June 2026 figure of "nearly 30" is right.

NEO (Tsinghua / Neuracle)37

Neuralink21

BrainGate16

Synchron10

Pittsburgh and Chicago9

Beinao-1 (NeuCyber)7

Caltech6

Georgia Tech (Kennedy)4

EPFL / Onward3

StairMed3

UCSF3

UMC Utrecht3

Grenoble / Clinatec2

Figure 5

Every brain-implant programme with 2 or more people, plus the newest ones

Each bar runs from the programme's first long-term implant to its last active year (or today). Dots mark records and milestones; hover for details.

University and hospitalSynchronNeuralinkChinaParadromicsRecord or firstTrial, approval or count

Georgia Tech (Kennedy)4 peopleFirst long-term implant

BrainGate16 peopleHandwriting

Pittsburgh and Chicago9 people

Caltech6 people

UMC Utrecht3 peopleFirst home use

Grenoble / Clinatec2 people

Synchron10 peopleUS trial starts

UCSF3 peopleSpeech 78 words/min

EPFL / Onward3 peopleWalking

NEO (Tsinghua / Neuracle)37 people32-person trial

Neuralink21 people12 people

Beinao-1 (NeuCyber)7 countedFirst 37

StairMed3 people

CorTec / Univ. of Washington1 person

Paradromics1 person10-min test

19982002200620102014201820222026

07 Why do people get a brain chip?

Every person in the count was paralysed or had lost the ability to speak. Of the 138, 68 had a spinal cord or other injury, 21 had ALS or another motor neuron disease and 12 had a stroke. For 37 the reason is not published one by one. Women are rare: 11 of the first 67 (16%) ^[1].

Figure 6

Why people got a brain implant

All 138 people counted. Neuralink (21), Beinao-1 (7) and 2 StairMed participants are in "not published": their trials take spinal cord injury, ALS or stroke, but the split is not public.

138people

• Spinal cord or other injury6849%
• ALS or other motor neuron disease2115%
• Stroke129%
• Not published per person3727%

08 How fast can a brain chip let someone talk?

The fastest rate in the papers we checked is 78 words a minute, decoded from a grid on the surface of the brain ^[24]; implants that go into the brain reached 62 ^[23]. Natural conversation runs at about 160 ^[23]. In everyday use one man with ALS spoke through his implant at about 32 words a minute with 97.5%% of words right, for more than 248 hours ^[25]. In 2016 the first system used at home managed two letters a minute ^[26].

Figure 7

How fast can a brain implant let someone communicate? (words per minute)

Best published rate for each kind of implant, against natural conversation. Handwriting is converted at 5 characters a word.

Natural conversation160

Speech decoding, surface grid (UCSF, 2023)78

Speech decoding, Utah arrays (Stanford, 2023)62

Everyday speech, Utah arrays (UC Davis, 2024)32

Imagined handwriting (Stanford, 2021)18

09 How long does a brain chip last?

Years, in some people. As of December 2023 the longest active user of a Utah array had had it for 8.5 years, and one early participant gave data for 13 years after the implant ^[1]. BrainGate’s 14 participants to 2021 kept their devices for 872 days on average ^[27]. Across all 67 people to 2023 the trials collected 2,380 months of data ^[1]. Signal quality varies: some arrays fade within a year, others keep working for four or more ^[1].

10 Can a brain chip make you smarter?

Not today, and no trial is trying. Of the 47 registered trials of long-term brain-reading implants we found, none accepts healthy volunteers and none uses an IQ test ^[3]. 5 track a short screening test of thinking (MoCA or MMSE), and only to check that the surgery did no harm. The implants read movement and speech signals from a few square millimetres of cortex: Neuralink’s has 1,024 electrodes ^[10] in a brain of about 86 billion neurons ^[31], roughly one electrode for every 84 million neurons.

Talk of brain chips that boost memory or link minds to AI is about future devices, not results so far. What the implants already do is give back abilities paralysis took away. For what everyday AI tools do to thinking, see AI and your brain; for the brain numbers that go viral, see our fact-check of 12 brain statistics.

11 Every programme in the count

Table 1

Every programme in the count (27 programmes, 138 people)

Sorted by people counted; click a column to re-sort. The CSV has one row per person.

Programme	Country	Device	First implant	People counted	Counted to	Source
NEO (Tsinghua / Neuracle)	China	Surface pads above the brain’s covering (epidural, wireless)	2023-10	37	Jul 2026	Tsinghua; ClinicalTrials.gov; BiopharmaTrend; Asia Economy
Neuralink	USA, Canada, UK, UAE (split not published)	Neuralink threads (N1: 1,024 electrodes on 64 threads)	2024-01	21	Jan 2026	AP; Neuralink; Bloomberg; Reuters
BrainGate	USA	Utah array (Neuroport)	2004	16	Dec 2023	Review (Patrick-Krueger et al. 2024)
Synchron	Australia (4), USA (6)	Stent electrode (Stentrode)	2019	10	Dec 2023	Review (Patrick-Krueger et al. 2024)
Pittsburgh and Chicago	USA	Utah array (Neuroport); surface grid (Cortac) for S1-S3	2011	9	Dec 2023	Review (Patrick-Krueger et al. 2024)
Beinao-1 (NeuCyber)	China	Surface electrodes (semi-invasive)	2025 (b	7	Mar 2026	Yicai; Reuters
Caltech	USA	Utah array (Neuroport)	2013	6	Dec 2023	Review (Patrick-Krueger et al. 2024)
Georgia Tech (Kennedy)	USA	Neurotrophic electrode	1998	4	Dec 2023	Review (Patrick-Krueger et al. 2024)
EPFL / Onward	Switzerland	Surface grid (WIMAGINE)	2021	3	Sep 2024	Review (Patrick-Krueger et al. 2024); ONWARD release
StairMed	China	Invasive wireless implant	2025-03	3	Apr 2026	Shanghai government; Pulse 2.0
UCSF	USA	Surface grid (Cortac)	2019	3	Dec 2023	Review (Patrick-Krueger et al. 2024)
UMC Utrecht	Netherlands	Surface grid (Medtronic, off-label)	2015	3	Dec 2023	Review (Patrick-Krueger et al. 2024)
Grenoble / Clinatec	France	Surface grid (WIMAGINE)	2017	2	Dec 2023	Review (Patrick-Krueger et al. 2024)
Battelle / Ohio State	USA	Utah array (Neuroport)	2014	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
Case Western	USA	Utah array (Neuroport)	2021	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
CorTec / Univ. of Washington	USA	Surface grid (Brain Interchange, records and stimulates)	2025-07	1	Oct 2025	IVAM
Feinstein Institutes	USA	Utah array (Neuroport)	2023	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
Johns Hopkins (movement)	USA	Utah array (Neuroport)	2019	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
Johns Hopkins (speech)	USA	Surface grid (Cortac)	2018	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
Paradromics	USA	High-density microelectrode array (Connexus)	2026-06	1	Jun 2026	Paradromics
Singapore NNI	Singapore	Utah array (Neuroport)	2019	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
TECNALIA (Spain)	Spain	Utah array (Neuroport)	2017	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
TU Munich	Germany	Utah array (Neuroport)	2022	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
Thomas Jefferson	USA	Utah array (Neuroport)	2020	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
University of Miami	USA	Surface grid (Medtronic)	2018	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
Wyss Center	Switzerland	Utah array (Neuroport)	2018	1	Dec 2023	Review (Patrick-Krueger et al. 2024)
Zhejiang University	China	Utah array (Neuroport)	2019	1	Dec 2023	Review (Patrick-Krueger et al. 2024)

12 How we counted

1. 01

   Start from the full review

   Patrick-Krueger, Burkhart and Contreras-Vidal searched PubMed, ClinicalTrials.gov, group websites and press releases and wrote to the research groups, and listed every person implanted in a long-term trial from 1998 to December 2023: 67 people in 21 groups ^[1]. Their Supplementary Table 2 gives each person’s code and trial ^[2]; each is one row of our CSV.

2. 02

   Add every later implant with a source

   For 2024 to October 2026 we took each programme’s own count: a company or hospital statement, a wire-service report or a trial registry. Where a programme gave several counts we used the latest exact one and kept vaguer later figures for the upper end of the range.

3. 03

   Check the registry

   We pulled every matching study from the ClinicalTrials.gov API (15 search terms, 29 October 2026) and kept the 47 trials of long-term brain-reading implants, to see who they enrol and what they measure ^[3]. Registry enrolment is not used as a head count except for NEO’s completed trial, whose 32 people match the company’s figure ^{[4] [6]}.

4. 04

   Keep only implants that read the brain

   Deep brain stimulators, cochlear implants and other devices that only stimulate are left out (there are about 736,900 cochlear implants alone ^[30]), and so are arrays placed for minutes during other surgery.

5. 05

   Name no one who has not gone public

   Each person appears under the code their trial uses. The one name in the data, Noland Arbaugh, is Neuralink’s first participant, who went public himself.