Filtra empresas com historico complete, calcula o acumulado, e pivota o resultado.

pivot_hive_table.hql

DROP TABLE IF EXISTS temp1;

CREATE TABLE temp1 (anomes INT, cnpj STRING, x DOUBLE);

INSERT INTO temp1 VALUES
  (201801, 'c0', 1),
  (201801, 'c1', 1),
  (201801, 'c2', 3),
  (201802, 'c0', 2),
  (201802, 'c1', 2),
  (201802, 'c2', 4),
  (201803, 'c0', 2),
  (201803, 'c1', 1),
  (201803, 'c2', 2),
  (201804, 'c0', null),
  (201804, 'c1', 2),
  (201804, 'c2', 5);
  
SELECT
  cnpj,
  collect_set(mes01)[0] AS mes01,
  collect_set(mes02)[0] AS mes02,
  collect_set(mes03)[0] AS mes03,
  collect_set(mes04)[0] AS mes04
FROM (
  SELECT
    cnpj,
    CASE WHEN anomes=201801 THEN cumx END AS mes01,
    CASE WHEN anomes=201802 THEN cumx END AS mes02,
    CASE WHEN anomes=201803 THEN cumx END AS mes03,
    CASE WHEN anomes=201804 THEN cumx END AS mes04
  FROM (
    SELECT
      cnpj,
      anomes,
      sum(x) OVER w AS cumx
    FROM (
      SELECT
        cnpj,
        anomes,
        x,
        count(x) OVER (PARTITION BY cnpj) AS countx
      FROM temp1
    ) t0
    WHERE countx = 4  -- filtra empresas com historico completo
    WINDOW w AS
      (
        PARTITION BY cnpj
        ORDER BY anomes
        ROWS BETWEEN UNBOUNDED PRECEDING AND CURRENT ROW
      )
  ) t1
) t2
GROUP BY cnpj;